* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download Identical Particles ( + problems 34
Thermodynamics wikipedia , lookup
Conservation of energy wikipedia , lookup
Photon polarization wikipedia , lookup
Classical mechanics wikipedia , lookup
State of matter wikipedia , lookup
Le Sage's theory of gravitation wikipedia , lookup
Quantum vacuum thruster wikipedia , lookup
Renormalization wikipedia , lookup
Introduction to gauge theory wikipedia , lookup
Bose–Einstein statistics wikipedia , lookup
Nuclear physics wikipedia , lookup
Internal energy wikipedia , lookup
Old quantum theory wikipedia , lookup
Van der Waals equation wikipedia , lookup
Fundamental interaction wikipedia , lookup
Relativistic quantum mechanics wikipedia , lookup
Density of states wikipedia , lookup
Grand Unified Theory wikipedia , lookup
Theoretical and experimental justification for the Schrödinger equation wikipedia , lookup
Standard Model wikipedia , lookup
Atomic theory wikipedia , lookup
Identical Particles Bosons and Fermions In Quantum Mechanics there is no difference between particles and fields. The objects which we refer to as “fields” in classical physics (electromagnetic field, field of distortions of an elastic medium, etc.) turn out to consist of quanta which to a certain extent can be considered as particles. The objects which we refer to as “particles” in the Classical Mechanics, can actually demonstrate a classical-field behavior. A typical example of the Quantum Mechanical field behavior is the interference phenomenon associated with the wave function. The wavefunction itself is not yet a classical field, because it is not directly observable. There is, however, a more deep property that relates quantum particles to fields. This property—the so-called exchange symmetry—is expressed by the axiom that the wavefunction of the system of identical particles should remain the same, up to a possible change of sign, with respect to interchange of the coordinates, ri and rj , of any two identical particles: Ψ(. . . , ri . . . , rj , . . .) = σ Ψ(. . . , rj . . . , ri , . . .) , σ = ±1 . (1) Clearly, the exchange symmetry implies the indistinguishability of identical particles, because now we cannot change the state of the particle i without identically changing states of all the other particles. For a given sort of particles, σ is always the same. Otherwise, the superposition principle of Quantum Mechanics would be violated. The particles for which σ = +1 are called bosons (after Bose), and the particles for which σ = −1 are called fermions (after Fermi). The sign of σ is of crucial importance for the properties of the particles. The properties of bosons and fermions are generally speaking radically different. The exchange symmetry reduces the number of allowed quantum states for identical particles. Suppose we have two non-interacting particles. Let ψa (r) and ψb (r) be the wavefunctions of two orthogonal single-particle states. Consider the following two very simple two-particle states: ΨI (r1 , r2 ) = ψa (r1 ) ψb (r2 ) , (2) ΨII (r1 , r2 ) = ψa (r2 ) ψb (r1 ) . (3) The physical meaning of the state ΨI is that the particle 1 is in the single-particle state a, while the particle 2 is in the single-particle state b. And ΨII is the same up to the permutation 1 ↔ 2. The states ΨI and ΨII do not feature the interchange symmetry. The only exception is the case a ≡ b, which is good for bosons. To construct proper bosonic and fermionic states, we need to (anti-)symmetrize the wave function. This means that if we want to deal with the two different single-particle states a and b for two identical particles, we have to write √ ΨBose (r1 , r2 ) = (1/ 2) [ψa (r1 ) ψb (r2 ) + ψa (r2 ) ψb (r1 )] , (4) √ ΨFermi (r1 , r2 ) = (1/ 2) [ψa (r1 ) ψb (r2 ) − ψa (r2 ) ψb (r1 )] . (5) This symmetrization procedure straightforwardly generalizes to any number of particles. For bosons we sum over all permutations of coordinates between different single-particle states. In the fermionic case, two or more equal single-particle states are not allowed, since this immediately leads, upon the anti-symmetrization, to zero wavefunction (Pauli exclusion principle). The crucial observation for both fermions and bosons is that for N identical particles and a given set of N single-particle states there is only one N -body state with the proper exchange symmetry that can be constructed out of the given set. Hence, a state of identical particles can be unambiguously specified by listing all the singleparticle states participating in the symmetrized wavefunction. And this is a very good news for the 1 Statistical Mechanics, since despite the entangled character of the eigenstate wavefunctions of identical particles, their nomenclature turns out to be quite simple. This nomenclature deals with the notion of occupation numbers. For example, if we have a box with periodic boundary conditions (PBC) so that the single-particle eigenstates are labelled by momenta (or wavevectors, which is basically one and the same), then each many-particle state of a system of identical particles is specified by corresponding set of occupation numbers {nk }, where each nk is a non-negative integer equal to the number of single-particle states ψk = (V )−1/2 eik·r participating in the N -body wavefunction. Though the set of the occupation numbers is infinite—because of the infinite number of possible wavevectors, only a finite number of them (≤ N ) is different from zero for a given N -particle state. By definition of the occupation numbers, X nk = N . (6) k For bosons each nk can assume any integer value from zero to infinity, while for fermions nk is either zero or unity. In terms of the occupation numbers, the eigenstate spectrum of a system of identical particles in a PBC box reads X E= ε k nk , (7) k where εk is the single-particle energy. The form of εk depends on what particles we are dealing with. For example, h̄2 k 2 for atoms, (8) εk = 2m εk = h̄ck for photons, (9) where m is the mass of the atom, and c is the velocity of light. Looking at Eq. (7), we note that if not for the constraint (6), each single-particle mode could be considered as an independent subsystem. We also see that in the case of bosons, where nk = 0, 1, 2, . . . each single-mode subsystem is equivalent to a quantum harmonic oscillator, while in the case of fermions, where nk = 0, 1 the single-mode subsystems are two-level systems. Later on we will show that in a macroscopical system the constraint (6) is actually equivalent to shifting the single-mode energies by some global constant µ (chemical potential). Meanwhile, we consider two particular cases: (i) photons/phonons and (ii) strongly degenerate fermi gas, in which the constraint (6) is either irrelevant from the very outset [case (i)], or can be easily taken into account by slightly changing the parametrization [case (ii)]. Equilibrium Photons and Phonons The constraint (6) is irrelevant to photons and phonons [phonons are the quanta of vibrations of liquids and solids; they are bosons] for a very simple reason: their total number is not conserved due to the processes of creation and annihilation. Correspondingly, the total number of photons/phonons is not an external parameter. Rather, it is just one of the internal thermodynamic characteristics of the system which essentially depends on temperature (and vanishes in the limit of T → 0). Hence, to find the thermodynamic properties of our system we can use the spectrum Eq. (7) and treat all nk ’s as independent quantum numbers. We then immediately have Z= Y Zk , (10) k where Zk = ∞ X e−εk nk /T = nk =0 2 1 1 − e−εk /T (11) is the partition function of a quantum harmonic oscillator—up to an insignificant global energy shift— with the energy quantum εk . By the standard thermodynamic procedure (line-by-line repeating the solution of the quantum harmonic oscillator problem) we get X F = T ³ ln 1 − e−εk /T ´ , (12) k S = X· k ¸ ³ ´ εk /T −εk /T 1 − e , − ln eεk /T − 1 X E = k µ CV = ∂E ∂T ¶ = εk ε /T k e − 1 , X (εk /T )2 eεk /T V (eεk /T − 1)2 k (13) (14) . (15) It is instructive to compare Eq. (14) with directly averaged Eq. (7): hEi = X εk n̄k , n̄k ≡ hnk i . (16) k By this comparison we conclude that the average occupation number is given by n̄k = 1 eεk /T −1 . (17) Problem 34. Derive Eq. (17) directly from Gibbs distribution by utilizing the trick ∞ X ne λn n=0 ∞ ∂ X λn = e . ∂λ n=0 (18) Employing a similar trick, find the statistical dispersion, ∆n, of the occupation number and analyze the relative dispersion, ∆n/n̄, in the two asymptotic cases: T ¿ ε and T À ε. Eqs. (10)-(15) are valid for any εk . Talking of photons, we are interested in the linear singleparticle dispersion law, Eq. (9). The same linear dispersion law applies to acoustic branch of phonons (with c the sound velocity), provided the temperature is low enough so that only the small-k linear part of the phonon spectrum is relevant. We also need to take into account that there are two different (transverse) polarizations of photons. Polarization can be treated as an internal state of a boson. Bosons (and fermions) with different internal quantum numbers behave like different sorts of particles. To account for the internal degrees of freedom we need only to multiply all the answers for additive quantities by the factor g equal to the number of different internal states. In the case of photons g = 2—what about phonons? In low-temperature fluids (like 4 He) there is only one (longitudinal) phonon branch, and thus g = 1. In a solid there are three branches: one longitudinal and two transverse. Generally speaking, the sound velocities of the longitudinal and transverse branches are different. This complication is not important for us here and from now on we assume that all the three velocities are equal. Replacing the summation over k with the integration, X k (. . .) → V (2π)3 3 Z dk (. . .) , (19) we get F = gT V (2π)3 Z ³ dk ln 1 − e−h̄ck/T ´ where C0 = − Z ∞ gT V 2π 2 = Z ∞ 0 0 ³ dk k 2 ln 1 − e−h̄ck/T ¡ dx x2 ln 1 − e−x ¢ = ´ = − π4 . 45 gC0 V T 4 , 2π 2 (h̄c)3 (20) (21) We arrive at an amazingly simple expression for the free energy: F = − gπ 2 V T 4 . 90(h̄c)3 (22) How does it turn out that the expression for the bunch of harmonic oscillators is much simpler than the expression for just one of them? It is a consequence of the so-called scale invariance of the problem. In the case of one oscillator of the frequency ω0 , there is a characteristic scale of energy, h̄ω0 , so that the answers can be written as some non-trivial functions of dimensionless variable T /h̄ω0 . In the case of the bunch of oscillators with the linear (or any other power law) dependence of εk on k, there is no any special energy scale. All the energy scales are similar to each other. The only possible function of temperature thus is a power law function. The physics of a single oscillator is totally absorbed by the dimensionless integral (21). In view of the simplicity of the expression (22) for the free energy, there is no reason in using generic relations (10)-(15). We readily obtain all the thermodynamic relations directly from (22): µ ∂F S=− ∂T ¶ CV = T ∂S ∂T ¶ µ = V µ P =− ∂F ∂V (23) gπ 2 V T 4 , 30(h̄c)3 (24) V E = F + TS = µ 2gπ 2 V T 3 , 45(h̄c)3 = ∂E ∂T ¶ = T ¶ = V 2gπ 2 V T 3 , 15(h̄c)3 gπ 2 T 4 . 90(h̄c)3 (25) (26) Problem 35. Make sure that Eqs. (23)-(25) agree with the generic relations (13)-(15). Hint: The dimensionless coefficients can be related to each other by doing corresponding integrals by parts. Spectral density. It is easy to note that the generic answers (12)-(15) for the thermodynamic quantities are actually depend on one function of one variable—the so-called spectral density, w(ε), which, up to a normalizing coefficient, is the “distribution function” for the energies of the harmonic oscillators: the quantity w(ε) dε is proportional to the number of oscillators (per unit volume) with the energies within the interval [ε, ε + dε]. To reveal this fact, we replace summation with integration, in accordance with (19), and then write—below F is an arbitrary function: Z dk F(εk /T ) ≡ (2π)3 Z dk (2π)3 Z dε δ(ε − εk ) F(εk /T ) . (27) That is we utilize the identity for the δ-function: Z ∞ −∞ dx δ(x − x0 ) = 1 . 4 (28) This trick allows us to make the replacement F(εk /T ) → F(ε/T ) and then change the order of integrations. We thus get Z Z dk F(ε /T ) = dε w(ε) F(ε/T ) , (29) k (2π)3 where Z dk δ(ε − εk ) . (2π)3 w(ε) = (30) Eq. (30) is the formal definition of the spectral density. If there are more than one branches of the bosons, we just sum up all the spectral densities. If the branches have one and the same dispersion law, this summation reduces to multiplying w(ε) by the factor g. For example, ε2 for photons . (31) w(ε) = 2 π (h̄c)3 Note the scale-invariance of this expression. Now we rewrite (12)-(15) in terms of w(ε). Z F = VT Z S = V ³ dε w(ε) ln 1 − e−ε/T ´ , · (32) ¸ ³ ´ ε/T dε w(ε) ε/T − ln 1 − e−ε/T , e −1 Z dε w(ε) ε , eε/T − 1 E = V Z CV = (V /T 2 ) dε w(ε) ε2 eε/T . (eε/T − 1)2 (33) (34) (35) We can also use (17) to find the total number of bosons by summing up all the occupation numbers: Z dε w(ε) . eε/T − 1 N = V (36) The form of Eqs. (34) and (36) is quite transparent: Z N ≡ V dε w(ε) nε , (37) dε w(ε) ε nε , (38) Z E ≡ V where nε = 1 eε/T −1 (39) is the average occupation number of the mode with energy ε. Clearly, the functions Wparticls (ε) = w(ε) nε and Wenergy (ε) = w(ε) ε nε play the roles of the distribution functions (over the single-particles energies, or frequencies—which is one and the same up to the Planck’s constant) for the particles and energy, respectively. This distribution functions should not be confused with the spectral density w(ε). The latter tells us only how the oscillators are “distributed,” saying nothing on whether they are excited or not. If w(ε) is a monotonically increasing function (normally this is the case up to some qualitatively irrelevant details), then the maxima of the distributions of energy and particles correspond to ε ∼ T , which is a very useful relation for order of magnitude estimates. 5 Problem 36. Give order of magnitude estimates for the following quantities: (a) Wavelength of the equilibrium EM radiation corresponding to the maximum of the energy distribution, at room temperature, (b) The same at the temperature of the Sun, (c) Temperature at which the maximum of the energy distribution of the equilibrium EM radiation corresponds to the red monochromatic light, (d) Pressure of the EM radiation corresponding to the case (c).—Compare this pressure to the atmospheric pressure. Degenerate Fermi Gas What is the distribution of the occupation numbers for fermions at T = 0? At T = 0 the system is supposed to be in its ground state. So, we need to minimize the energy (7) under the constraint (6). This constraint requires that N of the two-level systems be in their upper states anyway. Clearly, the minimum of energy corresponds to the case when the excited modes are just the first N minimal energies εp (in this section we label single-particle eigenstates by momentum rather than wave vector). We thus have: ( 1, if p ≤ pF np = (T = 0) , (40) 0, if p > pF where pF is called Fermi momentum. It is related to the total number of particles by the formula X 1=N . (41) |p|<pF In practice, this formula needs to be slightly corrected. The point is that fermions always have internal degrees of freedom because they feature half-odd-integer spin, and at least two internal states are guaranteed. Different internal states behave like different sorts of particles, which means that Eq. (41) should be written for each sort of the particles, with the value of N understood as the number of particles of the given sort. We will assume that all the internal states of our fermions are equivalent (basically, this means that there is no magnetic field), and that corresponding total numbers of particles equal to each others. If we want to reserve symbol N for the total number of particles, we need to replace N with N/g in Eq. (41). Using integration instead of summation, we get Z pF X V V p3 1= 2 3 dp p2 = 2 F3 . (42) N/g = 2π h̄ 0 6π h̄ p<pF That is à pF = 6π 2 g !1 3 h̄ n1/3 , (43) where n = N/V is the number-density of the gas. The single-particle energy corresponding to the Fermi-momentum is called Fermi-energy, εF = p2F . 2m (44) The total energy of the gas at T = 0 is E0 = g X p<pF p2 gV = 2 3 2m 2π h̄ Z pF 0 6 dp p2 p2 /2m = 3 εF N . 5 (45) With (43) taken into account, we have E0 = 3 (6π 2 /g)2/3 h̄2 10 m µ N V ¶2 3 N. (46) Differentiating with respect to volume, we find the equation of state µ ∂E0 P =− ∂V ¶ N (6π 2 /g)2/3 h̄2 = 5 m µ N V ¶5 3 (T = 0) . (47) Hence, the pressure of a Fermi gas is finite even at zero temperature and scales like n5/3 with density. Particles and holes. Consider now the case of finite, but low temperature, T ¿ εF . How can we excite our system? We can take some particle from an occupied state p1 < pF and place it into a free state p2 > pF . Actually, this way we produce two elementary excitations: a hole at the momentum p1 and a particle at the momentum p2 . The language of quasi-particles—particles with p > pF and holes with p < pF —is very constructive. It will allow us to easily describe the thermodynamics of a Fermi gas at T ¿ εF . First, we need to elaborate on the particle-hole language. We introduce the notion of Fermi sphere—the sphere in the single-particle momentum space defined by the condition p = pF . We treat the genuine particles which fill the ball under the Fermi sphere as vacuum, and our quasi-particles are the elementary excitations above this vacuum. From now on we use the word “particle” only for the states with p > pF . Particles and holes behave like anti-particles with respect to each other.—A particle-hole pair can annihilate. We introduce the occupation numbers for the quasi-particles: ( ñp = The constraint (6) then reads 1 − np , np , X if if ñp = X p ≤ pF , p > pF . ñp . (48) (49) p>pF p≤pF That is the numbers of particles and holes should coincide. We will see that in the limit T ¿ εF this constraint will be satisfied automatically, the total number of particles being fixed by the position of the Fermi momentum. This is the main advantage of using the quasi-particle language. We then introduce effective quasiparticle energy: ( ε̃p = εF − εp , εp − εF , if if p ≤ pF , p > pF , (50) which takes into account the fact that if we want to excite a single hole/particle we have to simultaneously put/take a genuine particle onto/from the Fermi surface. Now we can identically rewrite Eq. (7) in the form X ε̃p ñp . (51) E = E0 + p [We have also utilized Eq. (49).] At T ¿ εF the excited holes and particles will typically have momenta |p − pF | ¿ pF —we will also see it explicitly a posteriori—which allows us to use the approximate relation ε̃p ≈ vF |p − pF | , 7 (52) where vF = pF /m is the Fermi velocity. If we assume that all occupation numbers are independent parameters, we can easily find their average values, because these immediately follow from the statistics of the two level system. To find hñp i we need to consider a two-level system which first energy level is ε̃p · 0 = 0, and the second level is ε̃p · 1 = ε̃p . We then write hñp i = 0 · w0 + 1 · w1 = w1 , (53) where w0 and w1 are the probabilities to find our system in the lower and upper energy states, respectively. Hence, 1 hñp i = ε̃ /T . (54) p e +1 We are ready to calculate the numbers of particles and holes, Npart and Nhole . X Npart = g ñp = p>pF gV (2πh̄)3 Z p>pF dp ñp = gV 2π 2 h̄3 Z ∞ dp p2 pF evF (p−pF )/T + 1 . (55) Note that at T ¿ εF the exponential in the denominator of (55) will cut off the integration at p − pF ∼ T /vF ¿ pF . This means that we can replace p2 in the numerator with p2F . Then, introducing dimensionless variable x = vF (p − pF )/T , we get. Npart = gV pF mT 2π 2 h̄3 Z ∞ 0 dx 3 ln 2 T = N. +1 2 εF (56) ex Note that Npart /N ¿ 1. Analogously, Nhole = g X gV ñp = (2πh̄)3 Z gV dp ñp = 2 3 2π h̄ p≤pF Z pF dp p2 gV p2 ≈ 2 F3 +1 2π h̄ Z pF dp . +1 (57) With the same accuracy with which we replace p2 with p2F , we can formally set the lower limit of integration to −∞, since the actual cutoff of the integral corresponds to pF − P ∼ T /vF ¿ pF . Finally, introducing the new integration variable, x = vF (pF − p)/T , we get absolutely the same expression as Eq. (56). Hence, the constraint is satisfied. We could further proceed by standard thermodynamic algorithm: calculate partition function, free energy, and so forth. However, if we are interested only in the energy and heat capacity, we can skip these procedure by directly calculating the energy. Indeed, the average energy is related to the average occupation numbers by X E = E0 + g εp hñp i . (58) p≤pF 0 evF (pF −p)/T evF (pF −p)/T 0 p Utilizing, (54) and making the same approximations as when calculating the numbers of quasiparticles, we readily find X ε̃p hñp i = p V (2πh̄)3 Z dp ñp ε̃p ≈ V 2π 2 h̄3 It is known that Z ∞ dp p2 vF |p − pF | evF |p−pF |/T + 1 0 Z ∞ dx x 0 ex + 1 = ≈ π2 , 12 V pF mT 2 π 2 h̄3 Z ∞ dx x 0 ex + 1 . (59) (60) which leads to the final answer E = E0 + π2T 2 N 4εF (T ¿ εF ) . 8 (61) The heat capacity then is CV = π2 T N 2 εF (T ¿ εF ) . (62) Note that heat capacity per particle is much smaller than unity: As compared to the gas of distinguishable particles, Fermi gas at T ¿ εF has, in effect, much less excited degrees of freedom. In view of this fact, Fermi gas in this regime is called degenerate. Problem 37. Give an order of magnitude estimate for the Fermi energy of valence electrons in copper (in Kelvins). Comment: To a reasonably good approximation, valence electrons in metals can be treated as non-interacting Fermi gas [on the uniform background of ions—the so-called jelly model]. Give an order of magnitude estimate of the temperature T∗ (in Kelvins), at which the contributions to CV of copper from the phonon and electron subsystems are of the same order. Comment: In this temperature region, only the acoustic phonons are relevant. Problem 38. Calculate free energy of the degenerate Fermi gas (T ¿ εF ) and use it to calculate the finite temperature correction to the pressure. At what temperature this correction becomes comparable to the zero-temperature pressure? 9