You are on page 1of 30

# TFY4250/FY2045 Tillegg 8 - Tre-dimensjonal boks.

Ideelle Fermi- og Bose-gasser 1

TILLEGG 8

8. Three-dimensional box. Ideal Fermi
and Bose gases
This “Tillegg” starts with the three-dimensional box (8.1), which relevant e.g.
when we consider an ideal gas of fermions (8.2) and the ideal Bose gas (8.3).

8.1 Three-dimensional box
[Hemmer 5.2, Griffiths p 193, B&J p 331]

8.1.a Energy levels

The figure shows a three-dimensional box with volume V = Lx Ly Lz , where the potential
is zero inside the box and infinite outside. To find the energy eigenfunctions and the energy
levels, we use the well-known results for the one-dimensional box:
q
ψnx (x) = 2/Lx sin knx x; knx Lx = nx π, nx = 1, 2, · · · .

c ψ (x) = − h̄2 ∂ 2 h̄2 kn2 x π 2 h̄2 2
Hx nx ψn (x) = Enx ψnx (x), Enx = = n .
2m ∂x2 x 2m 2mL2x x
In the three-dimensional case, the Hamiltonian is H
c=H c +H
x
c +H
y
c , where all four
z
operators commute. As energy eigenfunctions we may then use the product states
s
8 nx πx ny πy nz πz
ψnx ny nz (x, y, z) = sin · sin · sin , (T8.1)
Lx Ly Lz Lx Ly Lz

which are simultaneous egenfunctions of H, c H c , H
x
c and H
y
c (and which are equal to
z
zero on the walls of the box). The energy eigenvalues are

π 2 h̄2 n2x n2y n2z
!
Enx ny nz = + + . (T8.2)
2m L2x L2y L2z

TFY4250/FY2045 Tillegg 8 - Tre-dimensjonal boks. Ideelle Fermi- og Bose-gasser 2

The eigenfunctions ψnx ny nz are normalized, and are also orthogonal, because we have only
one such function for each combination of quantum numbers nx , ny , nz . (Remember that two
eigenfunctions of a hermitian operator with different eigenvalues are in general orthogonal.)
Note that this means that the operators H, c Hc , H
x
c and H
y
c are a so-called complete set
z
of commuting operators.

If two or three of the lengths Lx , Ly and Lz are chosen to be equal, some of the excited
energy levels Enx ny nz will “coincide”, that is, some of the energy levels become degenerate.
As an example, with Lx = Ly 6= Lz , the states ψ211 and ψ121 get the same energy. With
Lx = Ly = Lz , we get even more degeneracy. (Try to find the degeneracy for the lowest
levels, and see B&J page 333.)
This shows that the degeneracy increases with increasing degree of symmetry. The same
effect is observed for the three-dimensional harmonic oscillator.

A small exercise: For a two-dimensional quadratic box you will find that the
energies are proportional to n2x + n2y . Use this to find the (degree of) degeneracy
for the three lowest-lying energy levels. [Answer: 2,1,2.] Are you able to find a
level with degeneracy 3? (Hint: Check the level with n2x + n2y = 50.)

8.1.c Density of states
An important application of the three-dimensional box is the quantum-mechanical descrip-
tion of ideal gases. We then consider a macroscopic volume V = Lx Ly Lz containing a
large number of identical particles (or large numbers of several particle species). The gas is
considered to be ideal, in the sense that we are neglecting possible interactions between the
particles.
For a macroscopic volume, the energy levels will be very closely spaced, and in the
statistical treatment of such many-particle systems the number of quantum states (wave
functions) per unit energy plays an important role. This is what is called the density of
states.
We have one spatial state ψnx ny nz (x, y, z) for each combination nx , ny , nz of positive
integers. Each such combination corresponds to a “unit cell” (with volume 1) in “n-space”.

TFY4250/FY2045 Tillegg 8 - Tre-dimensjonal boks. Ideelle Fermi- og Bose-gasser 3

π h̄ 2 2
Simplifying to a cubical volume V = L3 , so that E = 2mL 2 2 2
2 (nx + ny + nz ), we note that
the number Nsp (E) of spatial wave functions with energy less than E isto a very good
approximation given by the number of unit cells in the n-space volume in the figure, which
is 1/8 of a sphere with radius
s
q 2mEL2
|n| = n2x + n2y + n2z = .
π 2 h̄2
Thus the number of spatial states is
!3/2
1 4π 2 4π 2mEL2
Nsp (E) = · (nx + n2y + n2z )3/2 =
8 3 3 4π 2 h̄2
3/2
4π 2m

= V E 3/2 . (T8.3)
3 h2
We define the density of states, gsp (E), as the number of states per unit energy,

3/2
N (E + dE) − N (E) dN 2m

gsp (E) = = = 2π V E 1/2 . (T8.4)
dE dE h2

It can be shown (see e.g.Hemmer p 85) that this formula holds also for Lx 6= Ly 6= Lz . In
fact, it can be shown that it holds for rather arbitrary√forms of the volume V .
The fact that gsp (E) increases with the energy (∝ E) is characteristic for three dimen-
sions. If we go back to the one-dimensional box, the formula E = (π 2 h̄2 /2mL2 )n2 = (h2 /8mL2 )n2
tells us that the distance between the levels increases with the energy. Then the density of
states (the number of states per unit energy) must of course decrease. Explicitly, we have
that the number of eigenfunctions with energy less than E is
s
(1) 8mL2
Nsp (E) = n = E.
h2
Then the decrease of the density of states with increasing energy goes as
s
dN 2m
(1)
gsp (E) = = 2
· L · E −1/2 . (T8.5)
dE h

(T8. In such a case. a three-dimensional box (or potential well) with very small Lx . it follows in the same manner from the formula E = (h2 /8mL2 )(n2y + n2z ) that the number of states with energy less than E is (2) 1 2πmL2 Nsp (E) = · π(n2y + n2z ) = E.Tre-dimensjonal boks.7) dE h You should note that one.g. The x-dependent part of the wave function then is fixed to ψ1 (x) = q 2/Lx sin πx/Lx . This may imply that the particle (or particles) moving in this kind of well is in practice never excited to states with nx > 1. we see that the energy π 2 h̄2 /(2mL2x ) in the formula π 2 h̄2 2 π 2 h̄2 2 π 2 h̄2 2 Enx ny nz = n + n + n 2mL2x x 2mL2y y 2mL2z z becomes very much larger than π 2 h̄2 /(2mL2y ) and π 2 h̄2 /(2mL2z ). and the system is effectively two-dimensional. At the same time.og Bose-gasser 4 For particles that are confined to move in two dimensions (on an area A = L2 ). (2) dN 2πm gsp (E) = = 2 · L2 .and z-degrees of freedom.TFY4250/FY2045 Tillegg 8 . By choosing e. while Ly and Lz are large or even macroscopic. while the “longitudinal” dimension Lz is large (as shown in the figure on the right): . Thus the “physical processes will be going on in these directions”. “nothing happens” with the motion in the x-direction. only small amounts of energy are needed to excite the y.6) 4 h2 Here the density of states is constant. In a similar manner we can make a system effectively one-dimensional by choosing both Lx and Ly to be very small. (T8.or two-dimensional systems are not purely theoretical con- structions. Ideelle Fermi.

8. when we are not particaularly interested in boundary (or surface) effects. giving an effective one-dimensional system. 1 ψ(z) = √ eikz . Lz while replacing the “box conditions” by so-called periodic boundary conditions: ψ(0) = ψ(Lz ). the density of states. For the “one-dimensional” box above. the relevant condition is ψ(0) = ψ(Lz ). ψnz (z) = 2/Lz sin(πnz z/Lz ). but rather in bulk properties. we can motivate this statement by noting that the physics of this box will not be seriously altered (except for the disappearance of certain boundary effects) if we bend it around so that it is changed into a “ring”. For this ring.TFY4250/FY2045 Tillegg 8 . we may just as well use periodic boundary conditions to find e.g. and it is easy to see that they are also orthogonal (because they are momentum eigenfunctions. B&J p 331) For the effectively “one-dimensional” box just mentioned. the boundary effects mentioned are not very important for macroscopic Lz . while the transverse degrees of freedom are “frozen”. Lz Lz 2m These energy eigenfunctions are normalized.1. the use of periodic boundary conditions is as relevant as the original conditions. ±1. (T8. . =⇒ (T8.9) 0 In many problems.d Periodic boundary conditions (Hemmer p 86. Thus. each with their own momentum eigenvalue pn ): Z Lz ψn∗1 (z)ψn2 (z) dz = δn1 n2 . we have seen that the ordinary boundary conditions (the “box conditions” q ψ(0) = ψ(Lz ) = 0) give energy eigenfunctions in the form of standing waves. As explained in the references cited above. · · · . =⇒ eikLz = 1 =⇒ kLz = 2πn. it is often customary to replace these standing waves with running waves in the form of momentum eigenfunctions. However. Ideelle Fermi.8) 2πn nh h̄2 kn2 kn = . Griffiths p 199. For a three-dimensional box (modeling e. En = . Even if we do not “bend the box into a ring”.g. we may also here just as well use periodic boundary conditions. pn = h̄kn = . the three-dimensional well occupied by the conduction electrons in a piece of metal) it is impossible to imagine how to bend it around so that opposite sides could be “welded” together. ±2.og Bose-gasser 5 The energies corresponding to the lowest-lying (longitudinal) states ψnz (z) then become very small compared with the energies needed to excite the transverse degrees of freedom. n = 0.Tre-dimensjonal boks. This means that “intersting physics” may be going on “longitudinally” (involving the z-direction).

Tre-dimensjonal boks. and the same then holds for the density of states. Lx Lx and similarly for ky and kz . However. Lx Ly Lz V . nx = 0. Thus the number Nsp (E) of states with energy less than E is the same for both types of boundary conditions. px = h̄kx = nx . this is compensated by the former taking both positive and (1) negative values. corresponding to normalized wave functions and discrete momenta. so that (2π)3 (2π)3 3 d3 k = dkx dky dkz = dnx dny dnz = d n. periodic boundary conditions are even more relevant than box con- ditions. Then incoming and outgoing particles are most suitably represented by momentum wave functions.og Bose-gasser 6 In certain problems. in three-dimensional scattering calculations. A straightforward way to obtain the density of states is as follows: From the relations kx = 2πnx /Lx etc we have that dkx = 2π · dnx /Lx etc. One then uses periodic boundary conditions. as shown in the figure. in the form of a cubical box. Ideelle Fermi. Let us see how the periodic boundary conditions work in three dimensions. In one dimension we saw above that periodic boundary conditions give wave numbers kn = 2πn/Lz half as densely spaced (see figure (b) below) as the wave numbers knz = πnz /Lz obtained with ordinary box conditions (see figure (a) below). ±2. (T8. Such calculations are often simplified by placing the system inside a fictitious volume. ±1. It should be noted that periodic boundary conditions give the same density of states as the original box conditions. This is the case e.10) Lx Ly Lz where 2πnx h kx = .g. The orthonor- mal momentum eigenfunctions then are 1 ψnx ny nz = q eik·r . · · · .TFY4250/FY2045 Tillegg 8 .

1− v 2 /c2 (T8.13) dE cp v √ Armed with these formulae we first note that all states with energy less than E = c2 p2 + m2 c4 correspond to points in p-space inside a sphere with radius p.11) (2π)3 h3 This formula (and its generalization in the footnote) 1 is a common starting point in the calculation of densities of states. (T8. Ideelle Fermi.Tre-dimensjonal boks. From the above relation between d3 k and d3 n we may then state that the number of momentum states in the volume element d3 k of k-space (i.10) and (T8. p = γmv. According to (T8.TFY4250/FY2045 Tillegg 8 . h3 This formula is easy to remember because Each spatial stete occupies a “volume” h3 in phase space. E 2 = c2 p 2 + m 2 c4 . Let us try to calculate relativistically.og Bose-gasser 7 n-space p-space n-space now includes all eight octants (contrary to the octant on page 2).e. dvs. (T8.11) above may be generalized to the following rule: The number of spatial states in the element d3 rd3 p of the 6-dimensional phase space is d3 rd3 p dN = . . = 2 = .12) 2 2 cp c γmv = = v. E γmc2 From the last two formule we have that dp E 1 2E dE = c2 2p dp.11) are valid relativistically. both non-relativistically and in relativistic calculations. the number in d3 p in p-space) is V d3 k V d3 p dNsp = d3 n = = . using the formulae mc2 E=q ≡ γmc2 .11) the num- 1 The formula (T8. In this space we have one state per unit volume. The formulae (T8. The number of states in the volume element d3 n in this space thus equals d3 n.

17) h c This formula plays an importantrole in the derivation of Planck’s law. as we shall see later. that is.  with momentum < p. p + dp). (T8. h The relativistic formula for the density of states then becomes (using (T8. The relativistic formulae above hold not only for particles. we find that the number of photon modes with frequency in the interval (ν. .15)   dE h v relativistically.og Bose-gasser 8 ber then is   3 Number of states (Rel) V 4πp Nsp (E) = 3 .4):   3/2 Density of 2m  gsp (E) = 2π V E 1/2 . E + dE). that is for de Brogle waves.13))   2 density of Rel dN 4πV p gsp (E) = = 3 . ν + dν) is V · 4πp2 dp 8πV dNph = 2 3 = 3 ν 2 dν. for which the periodic boundary conditions give exactly the same allowed values as above for the wave number k (and the momentum p = h̄k of the photons). Ideelle Fermi. correspond to points inside a spherical shell in p-space: (Rel) V dNsp (E) = 3 · 4πp2 dp. States with energies in the interval (E. spatial states.  (T8. When we include an extra factor 2 to account for the two possible polarizations for each photon.  spatial states. (T8.3). Photons have v = c and p = E/c = hν/c. but also for electromagnetic waves.TFY4250/FY2045 Tillegg 8 .  (T8.14) This is the relativistic generalization of (T8.16)   h2  non-relativistically. √ In the non-relativistic limit (with p2 /v = mp = m 2mE) this formula yields equation (T8.Tre-dimensjonal boks. with momenta in the interval (p.    h 3 relativistically.

12 fermions. py . ny /Ly . This means of course that when all the N particles have been added. However.21 fermions inside the box volume V = Lx Ly Lz .TFY4250/FY2045 Tillegg 8 . The first few particles will then choose the spatial one-particle states ψnx ny nz with the lowest quantum numbers.a Fermi gas at low temperatures. If these particles were bosons. (Thus the only forces acting are those from the cofining walls. two particles with opposite spins in each spatial state. Ideelle Fermi. pz } = h{nx /Lx . However. the ground state of the system would correspond to all of the particles being in the box state ψ111 ∝ sin(πx/Lx ) sin(πy/Ly ) sin(πz/Lz ). that is. some of the properties of real systems are fairly well described by the idal-gas approximation. The radius pF of this sphere is called the Fermi momentum and the corresponding kinetic energy (EF = p2F /2m in the non- relativistic case) is the so-called Fermi energy. and real fermion systems are not ideal Fermi gases. and imagine that we add the particles one after one.g. they will occupy all states inside a sphere in p-space. the ground state of this many- particle system. This holds even for the ground state of the many-particle system. However. which is the state with the lowest possible total energy. and is an excited state of the many-particle system.og Bose-gasser 9 8. represented by points p = {px . as more and more particles enter. This is the state with the lowest possible total energy. generalities In the ideal-gas approximation we assume that there are no forces acting between the N identical spin. that is. since they are spin.2 Ideal gas of spin. nz /Lz } close to the origin in p-space. With a large number N of particles in the volume V . the value of |p| for the lowest-lying unoccupied state becomes larger and larger. Griffiths p 193. the conduction electrons in a metal. Let us use the one-particle states corresponding to periodic boundary conditions.12 fermions (Hemmer p 195. (A state with one or more unoccupied states inside this sphere and the same number of occupied states outside has a higher total energy. It is easy to find this ground state.2. To find pF .Tre-dimensjonal boks. this means that most of them are forced to occupy one-particle states with large quantum numbers. we only have to remember that . Later we shall use this ideal model to desribe essential properties of e. with all states inside occupied and all states outside unoccupied. B&J p 478) 8.) Real gases are not ideal.) Thus the N fermions in the volume V have momenta |p| ranging from zero up to a maximal momentum pF given by the radius of this sphere. Pauli’s exclusion principle allows only one particle in each quantum-mechanical one-particle state.

19) h3 3 Thus we see that even in the ground state of this many-particle system the particles have momenta and kinetic energies ranging from zero up to maximal values given by the Fermi momentum and the Fermi energy. (T8. Thus. If we set V = L3 . the Fermi energy in box 1 and box 2 are the same. which are h̄2 1/3 2/3 2N p2F N   pF = h̄ 3π og EF = = 3π 2 . This behaviour of the Fermi gas is very important. Ideelle Fermi. this effect of the exclusion principle is most pronounced for the lightest particles. each of the one-particle energies and hence also EF and Etot increase as 1/L2 for decreasing L. and so does the average kinetic energy h E i.Tre-dimensjonal boks. imagine that we keep V constant and increase N . say by a factor 8 = 23 . First of all we note that it is the number density N/V of the fermions that matters. since EF is proportional to V −2/3 . e. This illustrates the fact that the bulk properties of the Fermi gas are independent of the form and size of the macroscopic volume (and of the boundary conditions used).20) V 2m 2m V This property of the Fermi gas has far-reaching consequences. Equating the particle number N with the total number of states inside the Fermi sphere. (T8. the (non-relativistic) Fermi energy is inversely proportional to the mass. due to the exclusion principle. Then. Thus.18) h3 where the factor 2 counts the two spin directions. the Fermi energy still is the same. in white dwarfs. the radius pF of the Fermi sphere increases with a factor 2. Next. Note also that for a given N/V .og Bose-gasser 10 the number of states in the element d3 p of p-space is now V d3 p dN = 2 dNsp = 2 . and if we “merge” box 1 and box 2 as in box 3.TFY4250/FY2045 Tillegg 8 . (T8.g. and the Fermi energy EF = p2F /2m increases with a factor 4. h3 2m 0 3 |h {z 3 } 5 |{z} 2m N EF . we then have V 4πp3F N =2 . Total kinetic energy Etot and the average h E i (non-relativistic) To find the total kinetic energy Etot in the ground state of this N -fermion-system we only need to sum up the energies of the one-particle states inside the Fermi sphere: Z p2Z pF V d3 p  3 2  Etot = E dN = ·2 d p = 4πp dp 0 2m h3 2V 4π Z pF 4 2V 4πp3F 3 p2F = p dp = . The same result is obtained by reducing the volume V .

the pressure Pcl = N kB T /V of a classical ideal gas goes to zero. T → 0. In this limit.24) 5 V 15m V To see what this means.] The Pauli (or quantum) pressure It is interesting to note that if we keep N constant and let V increase.682×10−4 eV/K is Boltzmann’s constant. The conclusion is that the Fermi gas exerts a pressure ∂Etot 2 Etot P =− = (T8.38066 × 10−23 J/K=0. which is the state with the lowest total energy for the 14 electrons? What is the average energy of these electrons in the ground state? [Answer: The highest one-particle energy is that of the state with n2x + n2y + n2z = 9.) This is contrary to the quantum . In average. but is spent doing a work P ∆V = −∆Etot “on the outside”. ∂V 3 3 V we then find that an infinitesimal increase ∆V of the volume corresponds to a change of the total ground-state energy by the amount ∂Etot 2 Etot ∆Etot = ∆V = − ∆V. (kB = 1. Inserting for Etot we find that π 4/3 h̄2 5/3 2N N  P = EF = 3 .Tre-dimensjonal boks. ∂V 3 V To find out what happens with the “lost” energy amount |∆Etot | = −∆Etot . (T8. which is considered here.] A small exercise: A box with volume V = L3 contains 14 electrons.TFY4250/FY2045 Tillegg 8 . because Etot goes as Etot = constant · V −2/3 : h̄2 2/3 3 N  Etot = N 3π 2 ≡ konstant · V −2/3 . This means that |∆Etot | is not lost.22) 5 R EF [You should perhaps check that the integral Etot = 0 E g(E) dE. with g(E) = 2gsp (E).857.og Bose-gasser 11 Thus the total kinetic energy is 3 Etot = N EF .23) ∂V 3 V on the walls of the box. What is the highest one-particle energy (the Fermi energy) when the system is in the ground state. (T8. the total internal energy decreases. we note that the ground state. we must realize that the Fermi gas exerts a pressure P on the walls of the volume. we have n2x + n2y + n2z = 6. strictly speaking describes the system in the zero-temperature limit. Ideelle Fermi. 5 2m V With ∂Etot 2 2 Etot = − · constant · V −5/3 = − .21) 5 and the average kinetic energy of the N fermions is 3 hE i = EF . (T8. gives the same result.

the state of the gas does not differ much from the ground state: A few of the fermions have been excited. An even better name would be exclusion pressure. we must look into what happens with the state of the Fermi gas at “low” temperatures. (T8. Thus. this energy interval is centered around the energy value µ. At moderate temperatures. from states just inside the Fermi surface < < (EF − kB T ∼ E < EF ) to states just outside (EF < E ∼ EF + kB T ). quantum statistical mechanics (cf TFY???? Statistical Physics): At “low” temperatures.25) e(E−µ)/kB T + 1 distribution According to this distribution law the occupation number in general decreases with in- < creasing one-particle energy E. A state can be occupied one moment. that is. .TFY4250/FY2045 Tillegg 8 . because it is the exclusion principle which makes the Fermi sphere so large and the pressure so high. (Even in thermodynamic equi- librium there are fluctuations. To see this. namely some cen- tral results from another important physical theory. it gives the average number h n i of fermions occupiying the one-particle state in question. This probability (for a given one-particle state) is called the occupation number . when the “thermal” energy amount kB T is small com- pared to the Fermi energy EF . Ideelle Fermi. which does not go away as T → 0. this exclusion pressure is much larger than the pressure ex- pected from the classical ideal-gas formula P = N kT /V. The Fermi gas at “low” temperatures.og Bose-gasser 12 pressure we have found here. inside and close to the surface the probability of finding a state occupied is slightly smaller than 1. As shown in the figure. which is the E-value for which h n i equals 50 %. This pressure has nothing to do with thermal motion.Tre-dimensjonal boks. The Fermi–Dirac distribution To describe the state of this many-particle system of identical fermions at a temperauture T > 0 we have to invoke some results outside quantum mechanics. and is sometimes called the Pauli pressure. while it is slightly larger than zero for states outside and close to the surface. In statistical mechanics one learns that it is given by the so-called Fermi–Dirac- distribution: ! 1 Fermi–Dirac hni = .) According to the exclusion principle this occupation number must lie between zero and 1. We note also that the change from large (∼ 1) to small > (∼ 0) occupation numbers takes place in an energy interval of the order of a few kB T (cf the discussion above). and unoccupied in the next.

27) and we see that The Fermi–Dirac distribution predicts that the system is in the ground state at T = 0. the expected number of fermions in this interval is g(E)dE dN = h n i g(E)dE = . we see from the formula for h n i and from the figure above that the change from n = 1 to n = 0 happens very quickly: In this limit we have ( dN g(E) for E < µ(0). (ii) That the chemical potential µ depends on the temperature can be understood from the following argument: With g(E)dE states in the interval (E. Ideelle Fermi. m and N ). (iii) The fact that the difference between µ(T ) and µ(0) = EF is small for kB T << EF can be understood from the figure below: .Tre-dimensjonal boks. and also on the other system parameters (V. we may in practice set (T8.29) 0 e BT + 1 For given values of V and m the density of states g(E) is a well-defined function of the energy. Luckily. (T8. as it should.28) e(E−µ)/kB T +1 Integrated over all energies this should give the total number N of fermions in the volume V: Z ∞ g(E)dE N= (E−µ)/k . Then. What makes this law a bit difficult to grasp. kB T << EF . for T = 0 all states with E < µ(0) (and no states for E > µ(0)) are occupied. Thus. E + dE). for a given temperature T . we have µ = µ(T ). it can be stated that For “low” temperatures. the integral can only be equal to N for one particular value of µ.TFY4250/FY2045 Tillegg 8 .og Bose-gasser 13 This energy value µ is called the chemical potential for the fermion system in question. (T8. µ(0) = EF . But this was precisely how the Fermi energy was defined. The arguments behind this statement are as follows: (i) In the limit T → 0. (T8. lim = T →0 dE 0 for E > µ(0). is the fact that µ in principle depends on the temperature. which is thus a function of T . Thus.26) µ(T ) ≈ µ(0) = EF .

1/100 and 1/50. that is. and the expected number of fermions per energy unit. above and below the point A.TFY4250/FY2045 Tillegg 8 . we note that the area under the curve for T = 0 (that is. the same holds also for the areas under the two curves for T > 0. respectively. g(E) ∝ E . because g(E) decreases weakly as a function of the energy.Tre-dimensjonal boks. must be equal. .) This means that the two hatched areas in the figure below. These areas represent the number of unoccupied states inside the Fermi surface and the number of occupied states outside.) Thus. it must lie a little bit below the midpoint. (Actually. Since the two areas are equal. In the figure.og Bose-gasser 14 √ This graph shows the number of states per energy unit. dN g(E) = g(E) h n i = (E−µ)/k T . respectively. OF course. Point C lies “midway” on the “EF -line”.28). Ideelle Fermi. the value of dN/dE is 50 % of g(E). Here.) At the point B in the figure. we have exaggerated the difference between µ and EF . (Cf the integral (T8. and also the distance between A and C. under g(E) up to E = EF ) equals the number N of fermions. dE e B +1 for three cases: kB T /EF = 0. point A must lie almost midway on the “EF -line”. we understand that point A must in reality lie very close to point C. so the abscissa of B is the chemical potential µ.

dN .

.

g(EF ) .

= (E −µ)/k T ≈ 21 g(EF ). dE EF .

e F B +1 .

30) 12 EF Thus. as suggested. one can also understand that for kB T << EF the increase in the total energy Etot is small compared to the value at T = 0. we find the ratio P 2Etot /3V 2 EF = ≈ (kB T << EF ). 3 The Fermi temperature TF ≡ EF /kB depends on the system. but only to second order in the “smallness parameter” kB T /EF . with π 4/3 h̄2 5/3 N  P = 3 (T << EF /kB ≡ TF ) . Ideelle Fermi.Tre-dimensjonal boks. where this simple model gives some insight. (T8. the chemical potential µ decreases slightly with increasing temperature. When this ratio is small. It can be shown that the difference between µ(T ) and EF = µ(0) is of second order in kB T /EF :  !2  π2 kB T µ(T ) ≈ EF 1 −  (kB T << EF ). then almost all the fermions will be excited to one-particle states with high energies. It is the temperature for which kB T becomes equal to the Fermi energy EF . TF ≡ EF /kB .TFY4250/FY2045 Tillegg 8 . we realize that the occupation numbers can then very well become less than 50 % even for the “lowest” one-particle states (with E ≈ 0).25). (T8. Using Etot ≈ Etot (T = 0) = 53 N EF . T >> EF /kB . for most purposes. We have already mentioned the conduction electrons in metals. As stated above. It can be shown that   !2  2 2 π 2 kB T 3 5π 2 kB T Etot = Etot (0) + N = N EF 1 +  . 3 It turns out that the ideal Fermi-gas model can be applied with some success to several important physical systems.og Bose-gasser 15 This means that e(EF −µ/kB T ) ≈ 1. (T8. This means that also the increase of the total energy Etot becomes a second-order effect. The fraction of excited fermions (cf the hatched areas) is of the order of the smallness parameter kB T /EF . In all these cases.25). if the temperature is sufficiently high: If we heat our Fermi gas to extremely high temperatures. that is. This means that TF increases with increasing N/V . 2 From the graphs above. (T8. According to (T8. Since the number of available states is unlimited. EF − µ << kB T. . Anothe example is nuclear matter (in nuclei and neutron stars). that is. and the same can be said about the energy increase for these particles (∼ kB T ) compared to the average energy h E i.31) 4 EF 5 12 EF This means (among other things) that the exclusion pressure P = 2Etot /3V of the “low”- temperature Fermi gas is almost the same as for T = 0. 2 A small thougt experiment reveals that the chemical potential µ(T ) becomes considerably smaller than EF and perhaps even becomes negative. µ(T ) ≈ µ(0) = EF .33) 15m V Thus this relation is to a good approximation the equation of state for the Fermi gas when T is much smaller than the Fermi temperature. which is 53 N EF . negative. for which the model provides important insight. we can therefore use (T8. this means that the exclusion pressure is much higher than we would expect from the classical ideal-gas law Pcl = N kB T /V. µ(T ) must then be smaller than the lowest energies. the “thermal” part of the motion is almost negligible compared to the effect of the exclusion principle.32) Pkl N kB T /V 5 kB T The bottom line of this discussion is that the “cold” Fermi gas (with T << TF ≡ EF /kB ) behaves in several respects almost as in the ground state (for T = 0). and electrons in white-dwarf stars. it turns out that the actual temperatures are low compared to TF ≡ EF /kB .

.b Free-electron-gas model for conduction electrons in metals (Hemmer p 195.20) of the conduction electrons can be written as h̄2 h̄2 2/3 p2 N N  EF = F = 3π 2 = 2 (3π 2 a30 )2/3 . to 24. depending on whether the lattice of positive ions is that of a perfect or almost perfect crystal. we need to know the number Zl of conduction electrons per atom. that is. with the metal boundaries acting as high potential walls. forming a lattice of positive ions. they move around approximately as neutral free fermions. 2me 2me V 2me a0 V where h̄2 /2me a20 = 13.6022 × 1024 atoms per mole is Avogadro’s number. (T8. equal to a30 or thereabout. From this formula it is evident that EF must be of “atomic” size. Ideelle Fermi. These are the conduction electrons. the energies EF are of “atomic” size.og Bose-gasser 16 8. This gives us a simplified model in which the conduction electrons bahave as “netral” fermions. The table below gives N/V and the resulting Fermi energies EF for a selection of metals.TFY4250/FY2045 Tillegg 8 . Since mNa = ρm V. this periodic potential is replaced by a smoothed-out constant potential. In the free-electron model. Antallstettheter and Fermi-energier The Fermi energy (T8. as expected. we have that Na ρm ρm = = NA . B&J p 483) In a crystalline solid most of the electrons are bound to the nuclei at the lattice points. Griffiths p 193. because the volume V /N per conduction electron must be of the same size as the atomic volume. As you can see.6 eV is the Rydberg energy.34) V V A Using this formula one finds that N/V in metals varies from 0. which depends on the mass density ρm and the atomic mass. We also need the accurate number of atoms per unit volume. The actual potential seen by each conduction electron in a metal is a periodic or almost periodic potential in three dimensions. The latter is A m= . Na /V . A macroscopic potential well of this type is excellently represented by a three-dimensional box. moving without interacting with other particles inside the box.7 × 1022 per cm3 for beryllium. NA where A is the atomic weight (grams per mole) and NA = 0. we are treting the conduction electrons as an ideal Fermi gas. but if it is a metallic conductor some electrons from the outer subshells of the atoms are relatively free to move through the solid. V m A Thus the number density of conduction electrons is N Na Zl ρm = Zl = NA .Tre-dimensjonal boks.91 × 1022 perr cm3 for cesium. This number varies from 1 to 5 for the various elements. Because the charges of these electrons are cancelled by those of the ions. That is.2. For a more accurate calculation of N/V .

kB T ≈ 0.80 × 1010 N/m2 3 V 5 V .00 eV and the number density N/V = 8. Exclusion pressure and bulk modulus For copper the Fermi energy EF = 7.7 14.7 13.TFY4250/FY2045 Tillegg 8 . Solid State Physics.25 (From Ashcroft & Mermin. due to the large number densities N/V . Ideelle Fermi. This means that we have a “cold” Fermi gas of conduction electrons.47 × 1022 cm−3 corresponds to an exclusion pressure 2 Etot 2N P = ≈ EF = 2. which comes out typically as a number of the order of 10 eV.3 16.) The table also includes the Fermi temperatures TF = EF /kB .1 11.53 6.6 4.Tre-dimensjonal boks.90 5. the depth of the potential well mentioned earlier is the sum of EF and W .16 4.51 2. The work function is the energy required to liberate an electron at the Fermi surface from the metal. the latter are experimantal numbers.37 × 1029 eV/m3 = 3.22 Cu 1 8.40 2.86 5. The most striking feature of these results is that. of the order of 104 K.4 Ag 1 5.3 Be 2 24.3 Au 1 5.46 2.42 4.38 4. According to this model.og Bose-gasser 17 Element Zc N/V EF TF W (1022 cm−3 ) (eV) (104 K) (eV) Li 1 4. and the work functions W . the Fermi energies are much larger than the typical thermal energy kB T at room temperature (300 K).47 7.026 eV≈ 1/40 eV.74 5.92 Al 3 18.6 3.00 8. Another way to express the same thing is to note that the Fermi temperatures TF ≡ EF /kB are very large.49 6.38 K 1 1.12 2.70 4.

of the order of 4 × 105 atmospheres). 4 The heat capacity CV of the metal at normal temperatures is therefore due almost exclusively to the motion of the ions.TFY4250/FY2045 Tillegg 8 . this discrpancy between classical theory and experiments was a mystery. and the “Pauli repulsion” between the conduction electrons and the electronic cloud belonging to each ion. 4 Each of the “free” electrons have three translational degrees of freedom.og Bose-gasser 18 (which is a formidable pressure. then becomes large for the electron gas. This pressure corre- sponds to a small isothermal compressibility. κT ∂V T known as the bulk modulus. . ! 1 ∂V κT ≡ − . This is called the equipartition principle. [The stable ground state of the metal is a the result of a complicated interplay between the exclusion pressure. V ∂P T The inverse of of the compressibility. ! 1 ∂P Bel. which acts to expand the metal. the contribution from the conductione electrons is a factor 1 2 π kB T /EF π 2 kB T 2 = ∼ 10−2 3/2 3 EF smaller than expected classically. (T8. the attraction between the gas of conduction electrons and the positively charged ions. we get 5 Bel. which is 3N kB /2.] Thermal properties The free-electron-model result 2 2 π 2 kB T Etot ≈ Etot (T = 0) + N (kB T << EF ) 4 EF also gives a description of the specific heat of the electron gas.gas = P = 6. 3 This turns out to be of the same order of magnitude as the bulk modulus of the copper metal itself. which is measured to BCu = 13. In classical statistical mechanics. Before the advent of quantum mechanics. which comes out as ! ∂Etot 1 2 kB T CV = = N kB 2 π .35) ∂T EF This is much less than the classical result.gas ≡ = −V . one learns that the average kinetic energy for each of these degrees of freedom is 12 kB T . Ideelle Fermi.33 × 1010 N/m2 . Since P is propor- tional to V −5/3 .Tre-dimensjonal boks.4 × 1010 N/m2 .

Thus there is an energy minimum for a certain value of R. (T8. we then have a total number Nf = qNn of fermions with mass mf (me ) in the volume V = 43 πR3 . (See Hemmer p 209. is the exclusion pressure from the electron gas. Griffiths p 218. When this potential is replaced by the more realistic periodic potential mentioned above. As a function of R we see that the total energy Estar = Ekin + EG = b/R2 − a/R (T8. (There may also be other fermions present. 8.) To understand how this works we can use a simplified model (p 218 in Griffiths) where a star of total mass M is assumed to have constant mass density. dR R R 5 This section is not a part of the course. where Nion is the number of ions. Ideelle Fermi. with b= (qNn ) . but occur in energy bands with 2Nion states in each band. For large decreasing R it decreases (because EG = −a/R dominates).og Bose-gasser 19 Band theory of solids (Hemmer p 213. Assuming q fermions (electrons) per nucleon. and the star cools down. corresponding to a stable configuration. .c Electrons in a white dwarf star***5 (Hemmer p 196.2. (T8. Between these bands there are energy gaps with no states.36) R 4 A non-relativistic calculation then gives a total kinetic energy Ekin = 53 Nf EF which comes out inversely proportional to the square of the stellar radius R: h̄2 2/3 b 3 9π  5/3 Ekin = 2.39) is zero for R = ∞ . but the electron pressure will dominate.) With this band theory of solids it is possible to explain the difference between insulators.Tre-dimensjonal boks. the interior of the star is no longer able to withstand the “gravitational pressure”. (T8. metals and semiconductors. Band theory is a very important part of modern solid state physics. B&J p 484) When the fusion processes in an ordinary star die out.37) R 5 4 2mf It is easy to show that the total gravitational (potential) energy is Gm(r)dm Z Z R (ρ · 4πr3 /3)(ρ · 4πr2 )dr 3 GM 2 a EG = − = −G =− ≡− .TFY4250/FY2045 Tillegg 8 .38) r 0 r 5 R R where G is the gravitational constant. and the star may collapse to a white dwarf. because as we have seen. Griffiths p 198 and Brehm & Mullin p 596. This gives a Fermi momentum h̄ 9π pF = ( qNn )1/3 . What makes this process stop (if the stellar mass M is not too high). This value of R is given by dE 2b a = − 3 + 2 = 0. one finds a surprising result: The allowed energies are no longer distributed continuously. EF and P = 23 N V EF are inversely proportional to m. But when R decreases further the positive term Ekin = b/R2 will eventually make Estar increase again. The number of nucleons with mass mn then is Nn ≈ M/mn . Griffiths p 198) The main shortcoming of this free-electron-gas model lies in the assumption of a constant potential inside the crystal.

Ideelle Fermi.40) a 4 Gmf mn Nn1/3 2 By inserting numbers one finds that this simplified model gives a radius of the order of 102 –103 kilometers for a white dwarf with a mass of the order of the solar mass M.TFY4250/FY2045 Tillegg 8 . (T8.Tre-dimensjonal boks.og Bose-gasser 20 which gives a radius 2b 9π 2/3 q 5/3 h̄2 1   R= = .

With 1/3 1 9π 3 Gm2n 2   pF = qNn and EG = − Nn R 4 5 R (from the calculation above) we find that " 1/3 # 31 9π 4  E= h̄c q Nn4/3 − Gm2n Nn2 . More realistic models give somewhat larger radii. Heavier stars than this will therefore not end up as white dwarf stars. We therefore get a minimum in the energy E (and a stable star) only if the size of the kinetic term is larger than that of EG . . 5R 4 The main point here is that the kinetic term now has the same 1/R-behaviour as the potential term. To get an idea of what happens. This happens only if " 1/3 #3/2 h̄c 9π 4  Nn ≤ q ∼ 2 × 1057 . M ≈ 1. the exclusion pressure of the electron gas is unable to stop the gravitational collapse. who found the so-called Chandrasekhar limit.4(2q)2 M. This means that the behaviour of the function E = Ekin (R) + EG (R) will be modified for small R. with very high mass densities. Chandrasekhar in 1934. but still these white dwarfs are very compact objects. This was shown by S. In fact the density of the Fermi gas is so high that relativistic effects must be taken into account. (T8. making E positive for small R. we can use the ultra-relativistic energy-momentum relation E ≈ cp.41) Gm2n 4 The moral of this story is that if Nn is larger than this number.

of the order of 1044 nucleons per m3 . ∼ 1011 K. the exclusion pressure of the neutron gas will prevent further collapse. one finds that the radii of such neutron stars are a factor ∼ 2000 smaller than for a white dwarf. Although the neutron stars may have temperatures of the order of 109 K. they still are “cold” systems. e− + p → n + ν . only with mf = mn (instead of me ). because at these densities the Fermi temperatures are much higher. that is. the neutrons “go relativistic”. or 1015 g/cm3 . The density of neutron stars is then somewhat larger than for nuclei. The mass remaining in the core of such a supernova will create a neutron star. and the large energies released create an explosion known as a supernova. The mass limit is believed to be around 3M. and then the neutron star may collapse to a black hole. The reason is that at extremely high densities inverse beta decay. which are liberated and creates the explosion). Using the same formula as above for R. A heavier star than this will continue its collapse. if the mass is too big. only a few kilometers. But again. becomes energetically favourable. . The white dwarf radii for masses just below this limit turn out to be of the order of 5000 km. Provided that the mass of the neutron star is not too large. converting the electrons and protons into neutrons (and neutrinos.

It is also necessary to leave the ideal-gas approximation and take into account the interactions between the nucleons. . This limit requires detailed calculations involving general relativity theory. .

og Bose-gasser 21 8. R0 ≈ 1.07 fm and N/A ∼ Z/A ∼ 1/2 we find that these energies are approximately (p) (p) EF ≈ EF ≈ 43 MeV. 3 3 Thus each nucleon occupies roughly the same volume 43 πR03 in all nuclei. and one might be tempted to think of an almost static structure. To account for these facts it turns out that one has to assume that the nucleon-nucleon force is strongly repulsive for short distances. With V = A · 43 πR03 we find a Fermi momentum for the proton gas given by 1/3 1/3 2Z h̄ 9π Z   (p) pF = h̄ 3π = .5 fm. To get an idea about their motion we may use the simple Fermi gas model for nuclei. and is counterbalanced mainly by the exclusion pressures from the protons and neutrons.43) 2M R02 4 A 2M R02 4 A With M ≈ 939 MeV/c2 ≈ 940 me . (T8.4–0. or Brehm & Mullin.42) V R0 4 A and similarly for the neutrons.2.Tre-dimensjonal boks. as we know already from the uncertainty principle. 4 4 V = πR3 ≈ A · πR03 . M c2 .2 c.TFY4250/FY2045 Tillegg 8 . section 14-1. which acts as a spherical potential well of depth ∼ 50 MeV. This model treats the Z protons and the N neutrons as ideal Fermi gases contained within the nuclear volume. Brehm & Mullin p 695) The nucleons (protons and neutrons) in a nucleus are held together by the strong but very short-range nucleon-nucleon force. Ideelle Fermi. and the nuclear radius R goes as R ≈ R0 A1/3 . This attractive force is much stronger than the elec- tromagnetic repulsion between the protons. (A = Z + N ) (T8.07 fm.d Fermi gas model for nuclei (B&J p 510. The total binding energy of a nucleus is roughly proportional to the nucleon number A. This means that the nuclear volume is proportional to A. This shows that the nucleons have speeds of the order of s 2E q v=c ∼ c 1/20 ∼ 0. but of course such a picture is highly misleading: The nucleons are definitely not static. r ≤ 0. where R0 ≈ 1. (See B&J p 507. each ball representing a nucleon.) The almost constant density of nucleons sometimes make us picture the nucleus as a collection of close-packed “balls”. The two Fermi energies then are h̄2 h̄2 2/3 2/3 9π Z 9π N   (p) (n) EF = and EF = .

38).og Bose-gasser 22 so that it is reasonably correct to treat them non-relativistically. The figure shows isobars for A = 99.44) A A (See Brehm & Mullin pp 691–699.Tre-dimensjonal boks. This simplified model also explains why the most stable nuclei prefer to have N ∼ Z. (This holds mainly for light nueclei. but they are definitely moving around. it is relevant to keep the nucleon number A = Z + N fixed.TFY4250/FY2045 Tillegg 8 . and N = A − Z. Taken together. 5 5 with 3 h̄2 9π 2/3   k ≡ A· . we are neglecting all interactions between the nucleons. (T8. q. A A 3A 9A2 5/3 5/3 10x 20x2 ! N 2x   = ( 21 )5/3 1+ = ( 12 )5/3 1+ + + ··· . 5 2M R02 4 A nucleus can “transform” neutrons into protons or vice versa via β decay or inverse β decay.) Here. But that is due to the Coulomb repulsion between the protons. the calculation is analogous to the calculation of the gravitational energy in (T8. Also this term can be explained theoretically. the symmetry term and the Coulomb term predict that the most stable isobar (for a fixed A) will have N somewhat larger than Z. The second-order term in this expression goes as (9π)2/3 h̄2 ( 12 A − Z)2 . as found in our model. . which so far has been disregarded. 6 M R02 A and gives a qualitative explanation of the so-called symmetry term in the following em- pirical formula for atomic masses M (A X) = Z M (1 H) + (A − Z)Mn Z2 ( 12 A − Z)2 " # 2/3 − a1 A − a2 A − a3 1/3 − a4 + 5 /c2 . and using the binomial expansion.) Therefore. (See Brehm and Mullin. Introducing Z = 12 A − x and N = 12 A + x. An example is 238 92 U. p692 and p 761.e.d. we have 5/3 5/3 10x 20x2 ! Z 2x   = ( 21 )5/3 1− = ( 12 )5/3 1− + + ··· . For heavier ones there is a preference of larger N than Z. particularly for large A. while we vary Z and N = A − Z to find the minimum of the above expression for Ekin . Thus the simple model of two ideal Fermi gases gives a theoretical explanation of this empirical term. The third empirical term takes into account the Coulomb repulsion between the protons. A A 3A 9A2 It follows that 3 h̄2 1 20   (9π)2/3 Ekin ≈ 2 1 + ( 21 A − Z)2 . We see that the fourth term in this empirical formula is precisely of second order in 21 A − Z. It is easy to see that this sum is minimal for Z = N = A/2. 20 2M R0 A 9 which is minimal for Z = N = 12 A. Ideelle Fermi.) From the expressions above it follows that the total kinetic energy is 3 (p) 3 (n) Ekin = Z · EF + N · EF = k[(Z/A)5/3 + (N/A)5/3 ]. the square bracket represents the total binding energy of the nucleus.

For the most stable isobar. .Tre-dimensjonal boks. the Fermi levels for protons and neutrons then are at the same height. Ideelle Fermi. (See BM pp 692 and 761.) The Coulomb repusion term can be included into the fermi gas model by shifting the potential well for the protons upwards.og Bose-gasser 23 Note as mentioned above that a nucleus can transform into another isobar via β decay or inverse β decay. the simple Fermi model of the nucleus is able to explain at least qualitatively why the number of neutrons tends to exceed that of protons in the most stable nuclei. as shown in the figure below on the right. With this modification.TFY4250/FY2045 Tillegg 8 . In this situation the (kinetic) Fermi energy of the neutrons is higher than that of the protons. consistent with N being larger than Z.

is very important: It means. ! 1 Bose–Einstein hni = . For the massless photons the number is not conserved (it increases with T ).3 Ideal boson gas 8. the chemical potential µ(T ) is implicitly determined by the relation Z ∞ g(E)dE N= .Tre-dimensjonal boks. bosons in a way prefer to be in the same state. According to quantum statistical mechanics. Such a wave function can be written as a determinant. as in the corresponding relation (T8. and in statistical mechanics one then learns that µ = 0.) We also remember that for identical bosons (particles with spin 0. when a system of identical bosons is in equilibrium at a temperature T .46) 0 e(E−µ)/kB T −1 Here. 1. (T8. g(E) is the density of states. E + dE). (T8.og Bose-gasser 24 8. Here. For systems where the number N of bosons is conserved (as for example for a gas of helium-4 atoms in a volume V ). (See Hemmer p 189–192 or Griffiths p 179 for a reminder. and we remember that this determinant is equal to zero if we try to put more than one fermion into the same one-particle state. Brehm & Mullin p 552) Pauli’s exclusion principle is (as we remember) a consequence of the experimental fact that identical fermions (particles with spin 1/2.g. etc) Nature re- quires wave functions that are symmetric with respect to interchange of any pair of particles.47) ehν/kB T − 1 .TFY4250/FY2045 Tillegg 8 .29) for fermions. (photons) (T8.3. E + dE). We shall here briefly explain how this comes about. µ is again the chemical potential. however. the chemical potential will depend on the temperature (and on the other parameters). Ideelle Fermi.45) e(E−µ)/kB T − 1 distribution Formally. that the occupation number becomes larger than 1 for (E − µ)/kB T < ln 2. e. 5/2 etc) require wave functions which are antisymmetric with respect to the interchange of any pair of particle indices. Also for bosons. Then there is nothing stopping us from putting more than one boson into the same state.a The Bose–Einstein distribution (Griffiths p 214. 2. and the integrand is the expected number of bosons with energies in the interval (E.) Thus if we consider a wave mode with frequency ν and one of the two possible polarizations — or as we put it in quantum mechanics — a mode with energy E = hν — Bose–Einstein’s distribution law states that the expected number of photons (the occupation number) in this mode is 1 h n if = . the average (or expected) number of particles (occupation number) in a one-particle state with energy E is given by the Bose–Einstein distribution.. This sign. g(E)dE is the number of states in the energy interval (E. On the contrary. this formula differs from the the Fermi–Dirac formula only by sign of the last term in the denominator. 3/2. (See Griffiths p 216.

the occupation number 1 hni = (T8. (T8. so that all the occupation numbers are much smaller than 1.49) distribution law This is the Maxwell–Boltzmann distribution.3.TFY4250/FY2045 Tillegg 8 . The number of photons in each mode is given by (T8. (T8. it can be derived classically.og Bose-gasser 25 8.3. but first a little digression. The factor on the right is called a Boltzmann factor. and ! Maxwell–Boltzmann’s h n i ≈ e−(E−µ)/kB T . ν + dν) is 8πV dN = 3 ν 2 dν.50) h n ia favouring the state with the lowest energy. This means that the average number of particles in each state. we can all imagine that most of the particles get excited to high energies. Here we note the following important point: If we compare two one-particle states a and b. assuming that the particles are distinguishable.47). Then the particles can choose between a very large number of avalable states. Ideelle Fermi.52) c3 ehν/kB T − 1 .48) e(E−µ)/kB T ∓1 for most of the occupied levels tends to decrease with increasing temperature.51) c as was aalso shown p 8. The moral is that for high temperatures (or dilute gases). h3 With d3 p = 4πp2 dp and p = E/c = hν/c it follows that the number of modes in the frequency interval (ν. (T8. it doesn’t matter much whether the particles are fermions or bosons. Then the exponentials in the denominators are much larger than 1. multiplied by a factor 2 counting the two orthogonal polarizations: V d3 p dN = 2 . 8. (T8. and in a way is classical statistical mechanics in a nut-shell.b Maxwell–Boltzmann-fordelingen We shall return to the photons. The number of photons in the interval (ν. it follows that the ratio between the occupation numbers is h n ib = e−(Eb −Ea )/kB T .c Planck’s radiation law Let us now return to the photons. we may state that the number of electromagnetic modes in the phase space element V d3 p is given by (T8. ν + dν) therefore is 8πV ν 2 dν n(ν)dν = .Tre-dimensjonal boks. that is. His- torically the Maxwell–Boltzmann distribution was proposed long before quantum mechanics. when the system is at equilibrium at the tem- perature T .11). Because the electromagnetic waves in a cavity satisfy essentially the same boundary conditions as the de Broglie waves. If a gas of bosons or fermions is heated to very high temperatures.

These have been chosen such that atoms can “jump” between state 1 and 2 by emission or absorption of a photon with energy hν = E2 − E1 . we refer to the two states as state 1 and state 2. but when the system is in equilibrium. photons are constantly being absorbed and emitted by the walls. (See Hemmer p 9.og Bose-gasser 26 Multiplying with the photon energy hν and dividing with the volume V . we find the energy density (energy per unit volume) in the frequency range dν: ν 3 dν ! 8πh Planck’s u(ν)dν = 3 hν/k T . Even if the system is in equilibrium.53) c e B −1 radiation law This is Planck’s radiation law. Let us consider two particular states ψ1 and ψ2 for the atoms with energies E1 and E2 and let n1 and n2 be the number of atoms in the respective states. in equilibrium at temperature T .) 8. The numbers n1 and n2 will in general fluctuate. Brehm & Mullin p 170) The figure shows a box containing both photons and a number of atoms of a certain kind.54) n1 We also note that the energy density (per unit frequency) of the photons with frequency ν is given by 8πhν 3 1 8πhν 3 u(ν) = ≡ h n iph . in which Planck’s constant h was first introduced. (T8.Tre-dimensjonal boks. (T8. (T8. For simplicity.TFY4250/FY2045 Tillegg 8 . Even at equilibrium there is an ongoing “transport” of atoms up and down between the two states: Photons with frequency ν will excite some of the atoms from state 1 to state .d Einstein’s A and B coefficients (Griffiths p 311. the ratio between n2 and n1 is very accurately determined by the Boltzmann factor n2 = e−(E2 −E1 )/kB T = e−hν/kB T .55) c3 ehν/kB T − 1 c3 according to Planck’s law. Ideelle Fermi. and which in a way was the start of quantum mechanics. and also by atoms which are excited or de-excited.3.

To see how Einstein arrived at this result. we collect the three contributions. it even makes some of them emit radiation. The stimula- tion comes from the photons with frequency ν. and find that the rate of change of n2 is dn2 = n1 B↑ u(ν) − n2 (B↓ u(ν) + A). This emission is called sponta- neous because the probability is independent of u(ν). Solving for u(ν).og Bose-gasser 27 2. (T8. Therefore. At the same time there is a probability (per unit time) Psp. Thus we obtain the relation n1 B↑ u(ν) = n2 (B↓ u(ν) + A). who established this argument and introduced the A and B coefficients. Einstein had to assume that the probability (per unit time) of stimulated emission must be proportional to u(ν): Pst. and using the ratio for n2 /n1 given by the Boltzmann factor above. that is to the energy density u(ν): Pabs = B↑ u(ν). we have A 1 u(ν) = . B↓ (B↑ /B↓ )ehν/kB T − 1 . this rate can not counter balance the rate n1 Pabs .em = A of spontaneous emission. Therefore. dt At equilibrium. the so-called stimulated emission. This was a surprising result in 1917: Disturbing the atoms with radiation not only makes them absorb photons. Ideelle Fermi. showing that the three Einstein coefficients are not independent. This should give a rate n2 B↓ u(ν) of de-excited atoms. there- fore had to invent another contribution. The probability (per unit time) that an atom in state 1 absorbs a photon with frequency ν must be proportional to the chance to meet such a photon. which reduces n2 by a rate n2 A.Tre-dimensjonal boks.56) where = B↑ is a proportionality constant. except for fluctuations. (T8. both n1 and n2 must stay constant. The number excited per unit time then is n1 Pabs = n1 B↑ u(ν) : This rate adds to the population n2 in state 2.57) introducing a third coefficient B↓ . Einstein.TFY4250/FY2045 Tillegg 8 .em = B↓ u(ν).

og Bose-gasser 28 Still following Einstein we may campare with Planck’s radiation law given above. kR. Bu(ν).60) 3c where ω = 2πν.58) were discovered by Einstein already in 1917. B c h n if In this way Einstein was able to show that the probabilities (per unit time) for absorption and stimulated emission are equal. It turns out that the total probability (wem ) per unit time for emission into a mode which already contains n photons is given by a formula containing a proportionality . with hν ∼ 2 eV. Then hν/kB T ∼ 80. (T8. (T8.55). spontaneous emission is supplemented by stimulated or in- duced emission. Z d = ψ2∗ r ψ1 d3 r. is much smaller than 1. corresponding to λ >> R. ehν/kB T − 1 hν In such cases stimulated emission will dominate strongly over the spontaneous process. and the relevant occupation numbers h n if ≈ e−hν/kB T ∼ e−80 are so small that the probability of stimulated emission is non- existing. then we have hν/kB T ∼ 10−4 /0. From the second relation we note that the ratio between the probabilities for stimulated emission and spontaneous emission is equal to the occupation number: Pst. which requires that the photon wave number times the “size” R of the states. Einstein’s relations could be verified and the coefficients could be calculated (not only the ratios between them). and the occupation numbers become very large 1 kB T h n if = ≈ ∼ 250. if we consider a transition in the microwave region. (T8. This is reasonably well satisfied for most atomic transitions.58) A 8πhν 3 u(ν) coefficients = 3 = . (ii) On the other hand. ! the Einstein (T8. corresponding to emission of a photon in the visible region. Ideelle Fermi.59) Psp.025 eV) and consider an optical transition. The relations (T8.025 ∼ 1/250. Emission must then occur spontaneously.61) This result is valid in the so-called dipole approximation.TFY4250/FY2045 Tillegg 8 .Tre-dimensjonal boks.em Bu(ν) 1 = = h n iph = hν/k T . Quantum electrodynamics verifies Einstein’s results from 1917: When photons of fre- quency ν are already present. α is the fine-structure constant and d is the so-called dipole moment of the transition. while A/B comes out pro- portional to the third power of thefrequency: B↑ = B↓ ≡ B. After the development of quantum mechanics and quantum electrodynamics. The conclusion is that the two B coefficients have to be equal. It turns out that the spontaneous emission rate is 4ω 3 A = α 2 |d|2 . in (T8.em A e B −1 This result may be illustrated by two examples: (i) Suppose that we have an excited atom in a box at room temperature (kB T ∼ 0.

3. The probability for stimulated emission.) The question then is: Will the light beam be amplified or not? We first note that photons which are emitted spontaneously will be distributed isotropically. e B −1 which is otherwise typical in the container.og Bose-gasser 29 factor n + 1. which is the difference between these.62) in agreement with (T8. wem = (n + 1)wsp . this formula desribes spontaneous emission. we can conclude that the number of photons emitted “into” the beam will be smaller than the number absorbed. (T8. The beam will therefore loose intensity. Ideelle Fermi. and therefore give a negligible contribution to the beam.e Masers and lasers (B&J p 735. The photon modes (ki ) which are represented in such a beam can have much higher occupation numbers than the equilibrium value 1 h n iph = hν/k T . For n = 0. then becomes wstim ≡ wem − wsp = nwsp . Both these devices utilize stimulated emission from excited atoms or molecules. The photons (and other bosons) are “sociable”. such that n2 becomes larger than n1 . including the beam. Since the probabilities for absorption and stimulated emission are equal. the process happens much faster in the presence of suitable electromagnetic modes with large numbers of photons. 8. avsn. which is subjected to a beam of light. (The total system. The moral is that when an excited atom is ready to emit a photon. given a choice they will prefer the mode (state) with the largest occupation number n.59). LASER is the same thing with Light instead of Microwave. Let us imagine a container with atoms. . But this argument also tells us what is needed to get amplification: In some way or other we must bring the collection of atoms out of equilibrium. with a frequency ν corresponding to the energy difference between the states 1 and 2. This means that the total emission probability is a factor n + 1 times the spontaneous probability. 3-9 i Brehm & Mullin) MASER is an acronym for Microwave Amplification by Stimulated Emission of Radiation.TFY4250/FY2045 Tillegg 8 . and n2 normally is smaller than n1 .Tre-dimensjonal boks. This is called population inversion. is of course no longer in equilibrium.

one can obtain population inversion between levels 3 and 2.) 6 Pumping can be obtained for example by using electrical discharge or by using a lamp of sufficient intensity. This level is chosen such that it has a fast spontaneous decay to level 3. Ideelle Fermi. This way it is possible to produce highly monochromatic and coherent light. By a mirror arrangement one can obtain strong amplification in closely spaced optical modes with almost parallel wave vectors k. The figure illustrates the principle of the so-called four-level laser.Tre-dimensjonal boks. which is more long-lived. . with E1 < E2 < E3 < E4 .TFY4250/FY2045 Tillegg 8 . (See section 3-9 in Brehm & Mullin. If the decay from 2 to 1 is also fast. such that laser action with stimulated emission occurs between 3 and 2.og Bose-gasser 30 To make a laser or a maser therefore means to invent some mechanism to create pop- ulation inversion for two chosen levels. Using a suitable “pumping” mechanism6 one can raise the atoms from the ground state 1 to level 4.