CHAPTER 21 Applications of Second Quantization 1. Angular Momentum in a System of Identical Particles. An important ex- ample of an observable in a system of identical particles is the angular momentum operator, which according to (20.36) is the additive one-particle operator F = TEX Abwetsne ml I him) 1.) if the one-particle basis is characterized by the angular momentum quantum numbers j and m, as defined in Section 16.5, and « stands for all remaining quantum numbers needed to specify the basis. The total angular momentum operator (21.1) owes its simple structure to the absence of offdiagonal matrix elements of J with respect to j and «. The operator J? = J - F is not just the sum of the J? for the individual particles but contains terms that couple two particles, and thus it serves as an example of an additive two-particle operator. Since it conserves the number of particles, annihilating one and creating one, the operator ¥ commutes with the total number-of-particles operator, N. Exercise 21.1. Exhibit the two-particle matrix elements of the square of the total angular momentum explicitly. The two-particle states in which ¥, and J? have the sharp values MA and J(J + 1f® are readily constructed by the use of the Clebsch-Gordan coefficients defined in (16.67): Vr = CL hemes tamer isis |jaaoTM¥ (21.2) mime Since the one-particle state a!,,,{¥' is normalized to unity, (21.2) completely parallels expression (16.67), and the normalization constant C = 1, unless a = a, and j, = j.. The expression (21.2) remains an eigenvector of J, and J? even if a, = a and j, = jp, but the normalization is altered. The symmetry relation (16.78)526 Chapter 21 Applications of Second Quantization permits us to rewrite (21.2) in the form VS = AL (DPIC S aha moC dinars | JME (21.3) miymy with the upper sign applicable to bosons and the lower sign to fermions. Hence, the angular momenta of two identical bosons (fermions), which share all one-particle quantum numbers except m, cannot couple to a state for which J — 2j is an odd (even) number. If the usual connection between spin and statistics is assumed and bosons have integral spin, odd J values of the total angular momentum cannot occur for two alike bosons with the same « and j. Similarly, two identical fermions, which have half-integral spin, cannot couple to odd values of J if they have the same « and j. If ¥9}, ¥ 0, the value of the normalization constant C may be determined by requiring (¥'?},, ‘V'%},) = 1. The unitarity condition (16.76) yields readily the value C = 1/V2, so that if a) = a = « andj, =j, =/, 1 He fe Te Y Angered Sire | iJ M¥ (21.4) V2 mime (2) PA = Exercise 21.2. Verify the normalization (21.4). Exercise 21.3. Construct explicitly in terms of states of the form a},,,.4},,48° the total angular momentum eigenstates for two neutrons in the configura- tions (py2)* and (p3/2)®. How would the angular momentum eigenstates look if the two particles were a neutron and a proton but otherwise had the same quantum numbers as before? Exercise 21.4. Show that if two identical particles with the same quantum numbers « and with angular momentum j couple to zero total angular momentum, the resulting pair state is, in an obviously simplified notation, ¥8 = POs + Or? S (—1yalal FO 215) 2. Angular Momentum and Spin 4 Boson Operators. If we postulate a boson with spin 4 and with no other dynamical properties, the total angular momentum operator for a system of identical bosons of this kind takes the form f= Feel aya ) (21.6) 4-4/2, if the creation operators for the two spin states, m = +4 and —} are denoted§2 Angular Momentum and Spin § Boson Operators 527 simply by aj, and a‘ ,/.. Equation (21.6) may be decomposed into Ce t t Shaypaye, $= harpdyp h 21.7) Fr=5 (@hj2dy2 — a! 4/2012) : in agreement with our expectations for J ,. as the raising (lowering) operator which changes the state |JM) into |J M +1). From (21.7) we derive, using the boson commutation relations, he ae f= 4 (@iady2 + ahaa)? + ry (ated + 112412) a e N (N =n pvari(t 1) 21.8) 4 + 2 2 9 + (21.8) Hence, the state with a total number m = 2J of identical spin } bosons is an eigenstate of J? with eigenvalue J(J + 1)A®, where J is either integral or half-integral. The simultaneous eigenstates of the occupation number operators N,. = aj).4,,. (number of “spin up” bosons) and N_ = a! y.4_ 4. (number of “‘spin down”’ bosons) are simultaneous eigenstates of $? and #£. =(N, — N_)/2. The eigenvalues of the occupation number operators are determined by the relations n,tn=2J n,—n_=2M (21.9) or n=J+M n=J—-M Hence, by (20.34), the eigenstates, normalized to unity, are t toe (41)"* Ma" yp)? |JM) wo (21.10) V+ MIT — M)! In terms of the vector model of angular momentum, this representation of the state |JM) may be recognized as the projection MA of the resultant of 2 spin 4 vectors combined to produce the “stretched” vector polygon with all spin 3 vectors “parallel.” The requirements of Bose-Einstein statistics for the spins which make up this resultant cause this state to be uniquely defined. The violation of the connection between spin and statistics implied by the use of spin } bosons in this section does not vitiate the mathematical pro- cedure which we have outlined. The bosons defined here are not particles in the usual sense, since they have no momentum or energy. Rather, they are somewhat abstract carriers of spin, allowing a particularly transparent description of angular momentum states. As auxiliary entities, these bosons may be used for a relatively easy evaluation of the Clebsch-Gordan coeffi- cients and of the more complicated structures that arise in the coupling of528 Chapter 21 Applications of Second Quantization more than two angular momenta.’ An illustration of the usefulness of the spin 4 boson operator formalism will be found in the next section. 3. Elements of the Theory of Spin Waves and Magnons. The magnetic behavior insulators is frequently described in terms of an effective spin Hamiltonian, known as the Heisenberg Hamiltonian, which in the simplest case takes the form H = —YIsSyS, + Bebo 5 (Si). QL.) i In this expression S; is a spin operator corresponding to a localized spin (usually of an ion) at the site labeled by an integer j. The interaction con- stants J,, are assumed to be nonnegative numbers, appropriate to ferro- magnetic interactions which favor a lowering of the energy when the spins tend to be lined up parallel. Without loss of generality, we may assume I; = J,; and J,; = 0. (See Problem 4 at the end of this chapter for a deriva- tion of a Hamiltonian of the form (21.11) for spin } in a special example.) The second term on the right-hand side of equation (21.11) takes into account the effect of a uniform external magnetic field, directed along the positive z axis and of magnitude B, if gf, is the gyromagnetic ratio for the localized spins. We assume that all spins are identical and have maximum projection sh, where s may be integral or half-integral. For definiteness we also assume g > 0. It is easy to see by inspection of (21.11) that the ground state of the spin system is the state in which the z component of each spin is —sh and that the ground state energy is Ey = —> Js fh? — nBgBysh (21.12) i where n is the total number of spins or orbital sites occupied by spins. Assuming n to be an odd number, the summations may be extended from Zl=—(n— D/2 to @— 1/2. It is much less easy to determine the excited states of the system and the energy spectrum, especially if n is large, because all spins are coupled by the interaction, even if only neighboring spins interact. It is convenient to transform the problem by introducing the spin } boson representation of the preceding section. A distinct boson is defined for each spin site, and the appropriate site address j is affixed as subscript to the creation and annihila- tion operators. Operators pertaining to different sites are assumed to commute. e 1 For a full treatment, see J. Schwinger, On Angular Momentum in L. C. Biedenharn and H. Van Dam, editors, Quantum Theory of Angular Momentum, Academic Press, New York, 1965, p. 229.§3. Elements of the Theory of Spin Waves and Magnons 529 Equation (21.9) imposes a constraint on the boson operators: A j0,5A4y2,5 + A ay2,s0-ay2,5 = 28 (21.13) Using this condition and the relations (21.7) for each spin, we obtain for the scalar product of any two spins: Tt t + t S;S,= | Auer sA—ap2 yt Hanya 31/2, 581214220 + Ai syya,5 — S\(A4y2,:01/2,1 — 9] (21.14) The form of the Hamiltonian makes it evident that, in the low-lying excited States of the system, the spins remain pointed downward with high proba- bility and the excitations are due to small departures from this perfect downward polarization of all the spins. The approximation to be made is based on these considerations and is characteristic of quantal many-body theories. It assumes that, with regard to their effect on low-lying energy eigenstates of the system, the operators af), , and a,/.., ate “small” compared with the operators a 05 and a_,,.,and that, in the Hamiltonian, terms which are of order higher than second in aj,,, and a,.,, may be neglected if n is large. Using the expansion Aya. sAya.5 8s dictated by (21.13), we obtain from (21.14) and (21.11) the approximate Hamiltonian + G_yja,j © Aaya,s © [25 — H, app = Ey + 2hs z 2 Jin, s0r2.5 — 2h's & Jai strye4 + BgBoh ¥ aij, sAr0.5 (21.15) 5 Further progress toward solving the eigenvalue problem for this approxi- mate Hamiltonian can be made if the spin sites exhibit a periodicity such as is encountered in a crystalline solid. Since the essence of the method emerges clearly even in a very simple example, we shall consider only a one-dimen- sional periodic lattice in which all sites are equivalent so that Jisaua = In (21.16) If n, the number of spin sites, is large, mathematical simplicity is achieved, without any significant alteration of the energy spectrum, by ignoring end effects and taking the limit n + oo. In the infinite periodic one-dimensional lattice, equation (21.16) must hold for any integer g, and both and Happ are invariant under any translation of the spins from site j to site j + 4.530 Chapter 21 Applications of Second Quantization This symmetry suggests that the Hamiltonian be transformed from the site (or coordinate) representation to the displacement (or momentum) representation. The unitary operator D, which effects such a displacement is defined by the relations Dain, sD) = iyo, s40 (21.17) and by the property D{¥ =F, if ‘Y denotes the ground state of the system in which there are no “spin up” bosons. In analogy with the derivation of Section 14.7, we find that the operators b= YD aly sei? (21.18) which are defined in the interval 0 < x < 27 have the simple property Dyol Dt = ea (21.19) This equation shows that a,"¥ is a one-boson eigenstate of D, with eigenvalue e~‘**. Since D, is unitary, « must be real. Equation (21.18) defines a continuum of new creation operators af characterized by the variable x, and these operators may be used instead of the discrete set of operators aj), ,. Since “ala, de = ¥ al 21.20) Jy te =D ale sy.s (21.20) the transformation conserves the number of “‘spin ups” and is precisely of the form (20.12). If it is remembered that «, is defined only for 0 Ine? (21.23) is independent of /. Using the /attice sum (0 < x, x’ < 2m) +2 > eel ade — K') (21.24) 10 we finally obtain the Hamiltonian in terms of the magnon creation and annihilation operators: 2 F wy = Ey + [ dic ofag{2HsJ — 2KPsq(x) + BgBofi] (21.25) 0 Equation (21.25) is the desired result of our calculation and has a straight- forward interpretation. Since a!a, is the number density of magnons in «-Space, the quantity in brackets, E, = 2h*sJ — 2hsq() + BgBoh (21.26) is the energy of a magnon with displacement quantum number x. If d is the distance between adjacent spin lattice sites, «/d becomes the wave number, and p = hix/d may be identified as the linear momentum of a magnon. Its range is 0 < p < h/d, in harmony with the fact that the shortest wavelength sustained by the lattice is equal to the spacing d. Equation (21.26) is called the spin wave dispersion formula. Since # pp is expressible as a linear combination of occupation number operators ajo, the approximation made in deriving #,.)) from #% consists of neglecting any interactions between magnons. #,), describes noninter- acting magnons, and the excited eigenstates of the approximate Hamiltonian are simply the states obtained from the magnon vacuum ¥) (ground state of the system) by creating one or more magnons in states with definite values of x. The energy of such a state is the sum of the individual magnon532 Chapter 21 Applications of Second Quantization energies. The one-magnon eigenstate with definite « is of YO = Ly orig! WO (21.27) Qa i Exercise 21.5. Prove that (x) is an even function. Show that in the low- momentum limit the energy-momentum relation for a magnon is similar to that for a nonrelativistic particle. Identify the effective mass of a magnon. Exercise 21.6. Derive the spin wave dispersion relation for a lattice in which each spin interacts only with its nearest neighbor. Exercise 21.7. Prove that while (21.27) is an eigenstate of #, ,, the state 1 inj, t = Feat Aah N20 7 is an eigenstate of the exact Hamiltonian #, and calculate the corresponding eigenvalue if the spin lattice is one-dimensional and all spins are equivalent. Exercise 21.8. From the Hamiltonian (21.11) derive the equations of motion for the spin operators S, in the Heisenberg picture. Assuming that the operators may be replaced approximately by classical vectors and that, as is the case near the ground state, each S,(t) differs only slightly from a fixed common vector S, obtain linear equations of motion for the spin deviation vectors §,(t) = S,(t) — S, if |s,(t)| « |S]. Show that the equations of motion for the spin deviation vectors describe the propagation of spin waves through a periodic lattice, each vector s,(t) precessing about S. For a one-dimensional lattice in which all spins are equivalent, show that the relation between the precession frequency w and the wave number x/d is given by the dispersion formula (21.26), provided that the magnon energy is identified as E, = hw. 4. First-Order Perturbation Theory in Many-Body Systems. A simple and important illustration of the use of two-particle operators is afforded by a first-order perturbation calculation of the energy eigenvalues of a Hamiltonian which describes a system of interacting identical particles: H =D eaia, + 4D a) a} a,adqrl V Its) (21.28) ‘ e It is assumed that the eigenstates of the unperturbed Hamiltonian of non- interacting particles, Ho =D e010; (21.29)§4 First-Order Perturbation Theory in Many-Body Systems 533 are known and characterized as |m,,nz,...n;...) by the eigenvalues n, of the occupation number operators ata,. If the eigenvalues of #9 are non- degenerate, first-order perturbation theory gives for the energies the approx- imate values Enum = Lm + ED (ita ah al aya, |nns---Yqr| V|ts) (21.30) arst In evaluating the matrix element of the operator alata,a,, it is helpful to recognize that, owing to the orthogonality of the unperturbed eigenstates, nonvanishing contributions to the interaction energy are obtained only if q# rand either s=randt=qors=qandt=r, orifg=r=s=t. Equation (21.30) is therefore reducible to Enuny = Z mite + FS mareliarl V lar) & Carl V Igy] ¢ r + 4X ngng — 1)aal V aq) (21.31) The + sign holds for Bose-Einstein statistics and the — sign for Fermi-Dirac statistics. The two matrix elements (gr| V |gr) and (gr| V |rq), connecting the two one-particle states g and r, are said to have direct and exchange character, respectively. The last term, which accounts for the interaction of particles occupying the same one-particle state, vanishes for fermions. The evaluation of a matrix element of the product of several creation and annihilation operators carried out here is typical of most calculations in many-body theories. The labor involved in such computations is significantly reduced if the operators in a product are arranged in normal ordering, ie., with all annihilation operators standing to the right of all creation operators. The operators in the Hamiltonian (21.28) are already normally ordered. If a product is not yet normally ordered, it may, by repeated application of the commutation relations, be transformed into a sum of normally ordered products. A set of simple manipulative rules may be formulated? which permit the expansion of an operator of arbitrary complexity into terms with normal ordering. As an example, we briefly consider the fundamental problem of atomic Spectroscopy, the determination of energy eigenvalues and eigenstates of an atom with n electrons.‘ If all spin-dependent interactions are neglected, only 3 The operator algebra of normal ordering, and especially its relation to time ordering, was developed by G. C. Wick. For a lucid description of the use of these concepts and related diagrammatic techniques in the Brueckner-Goldstone theory of nuclear matter see B. D. Day, Rev. Mod. Phys. 39, 719 (1967). +A useful introduction to atomic, molecular, and solid state applications of quantum mechanics is M. Tinkham, Group Theory and Quantum Mechanics, McGraw-Hill Book Company, New York, 1964.534 Chapter 21 Applications of Second Quantization electrostatic potentials are effective. In this approximation both the total orbital and the total spin angular momentum commute with the Hamiltonian. As was suggested in Section 17.6, it is practical to require the eigenvectors of #,, on which the perturbation theory is based, to be also eigenvectors of the total orbital and the total spin angular momentum. A level with quantum numbers L and S is split by the spin-orbit interaction into a multiplet of eigenstates with definite J values ranging from |L — S| to L + S. This scheme of building approximate energy eigenstates for an atom is known as L-S (or Russell-Saunders) coupling. If #4 is a central force Hamiltonian for noninteracting particles, the unperturbed eigenstates are characterized by the set of occupation numbers for the one-particle states, or shells, or orbitals with radial and orbital quantum numbers n,, /;. Such a set of occupation numbers is said to define a configuration. A particular configuration usually contains many distinct states of the product form t © TH en,acmm, F Eigenstates of # that are represented by a product of n creation operators are said to be independent particle states. Although generally knowledge of the atomic configuration and the quantum numbers L and S, M;, and Mg is not sufficient to specify the state of an atom unambiguously, in simple cases, e.g., near closed shells, these specifications may determine the state uniquely. The states of the helium atom may be fully classified in this way, and we shall discuss these in some detail. If the two electrons are in different shells, the states of any two-electron configuration (m,/,)(m2/,) which are simultaneously eigenstates of the total orbital and the total spin angular momentum are, according to the formula (21.2) Waitingt(LSMzMs) = > (lylamym, | haLM;) x. adm; mg | HESMs) mma x al apo) (21.32) malaga" 2mm If the two electrons are in the same shell and the configuration is (n/)?, it is legitimate to set n, = n, =n and /, = /, =/ in (21.32) provided that a normalization factor of tv is furnished. The ground state of the neutral helium atom is described by the configura- tion (1s)? and has the spectroscopic character Sy. In our notation this state may be expressed as Y¥{$10(0000) = ajoo,_s/2@00,1242 (21.33)§5 The Hartree-Fock Method 535 The configuration of the excited states is (1s)(n/) with n > 1. Since the two spins may couple to 0 or 1, the excited states are classified as singlet (S = 0) and triplet (S = 1) states. With the appropriate values for the Clebsch- Gordan coefficients substituted in (21.32), we obtain for the triplet states: Wiu(L Lm, 1) = ahinr/2Ao0/8h Pieu(1 1 m,0) = 5 him -aetnae + im,r/2A0,-2)¥" (21.34) v2 Wiel m, A) = Gham,1/24I00 108 and for the singlet states: we (10m0) = (at fo0./2 — Aram,aledieo,-) EO — (21.35 10n(/0m0) = fa Grim satioaan n1m,1/24100,-1/2) (21.35) Owing to the anticommutation properties of the creation operators, the triplet states are symmetric under an exchange of the spin quantum numbers of the two particles and antisymmetric under exchange of the set of radial and orbital quantum numbers. The situation is reversed for the singlet states. The perturbation interaction, arising from the Coulomb repulsion of the electrons, is diagonal with respect to all the unperturbed states which we have constructed, and the first-order corrections to the energy are the expectation values of the interaction in these states. These energies were already worked out in terms of direct and exchange integrals in Section 17.9. We now see that the identity of the electrons, manifested in their statistics, results in a definite correlation between the spatial, or orbital, symmetry and the total spin S of the system. The states of parahelium are singlet states, and the states of orthohelium are triplet states. In complex atoms the connection between S and the spatial symmetry of the state is less simple and not necessarily unique, but S remains instrumental in classifying the orbital symmetry of the states and thus serves as a quantum number on which the energy levels depend, even though the interaction depends only on the position coordinates of the electrons.® 5. The Hartree-Fock Method. One of the most useful methods for approxi- mating the ground state of a system of n interacting fermions is based on the variational property of the Hamiltonian H => bial Hy la’ybg + 4 Y bbs aBl V la’B’yby-by (21.36) aa’ apa’ B’ 5 For acompact treatment of the theory of atomic spectra in terms of the second quantiza- tion formalism, see B. R. Judd, Second Quantization and Atomic Spectroscopy, Johns Hopkins Press, Baltimore, 1967.536 Chapter 21 Applications of Second Quantization The essence of the Hartree-Fock method is to seek a new one-particle basis with creation operators aj such that the individual particle state + aate (21.37) renders the expectation value of # stationary. In this new basis the Hamiltonian appears as # = ak| Ho |a, +4 Y alaigrl V |ts)a,a, (21.38) fs ou The exact ground state of the Hamiltonian is, of course, usually not as simple as ¥, but can be thought of as a linear combination of individual particle states, with the expression (21.37) as the leading term. The variation to be considered is a basis change, which is a unitary transformation and expressible as a, + bay = > 45 (5, + ies) (21.39) or : bag = 1S aes a with transformation coefficients ¢,, such that lenl «1 The general variation of the state Y’, can be built up as a linear combination of independent variations of the form OV, = ey,a'a,¥, (21.40) where a, must annihilate a fermion in one of the occupied one-particle states 1---n, and af must create a particle in one of the previously unoccupied one-particle states n+1,.... The unitarity of the transformation coeffi- cients in (21.39) requires only that the ¢,, form a Hermitian matrix. Since the variation 6‘¥,; vanishes owing to the exclusion principle, the condition &3 = &,* implies no restriction on the permissible variations, and the independence of the e-variations is assured. (Variations with j = k do not change the state and are therefore irrelevant.) We may thus confine our attention to variations 6¥ of the form (21.40) which are orthogonal to the “best” state ’, of the structure (21.37). The variational theorem, 5(#) =0 in conjunction with the Hermitian property of #, requires that (FL, FOV, HY.) — (Fy, HOF OF, ¥,) = 0 The orthogonality of 6¥ and ¥, guarantees that the variation preserves the§5 The Hartree-Fock Method 537 normalization of the state and, according to the last equation, makes it necessary that 6 also be orthogonal to #¥,. Hence, the variational condition is (Brillouin’s theorem) (OF, #¥,) =0 (21.41) If the Hamiltonian (21.38) and the variation (21.40) are substituted into this condition, we obtain (a}ay¥o, E anaxnl Hy |Y,) + (aay S aata,adarl V [ts)¥,) = 0 ml arst This relation is easily seen to be equivalent to the equation 2 (jl Ho lk) + > [itl V |kt) — Cit] V [tk] = 0 (21.42) rat The sum over t is to be taken only over the occupied one-particle states; j denotes any unoccupied, k any occupied one-particle state. If in the original one-particle basis b{ the interaction between the fermions is diagonal, (Bl V |a'B') = Veg Sax: Spor (21.43) condition (21.42) can be construed as expressing the orthogonality of the eigenkets of an effective one-particle Hamiltonian, Hj, corresponding to the eigenvalue problem Hae We) = [Ho + SS Ce | BVelB | 22 — aye | B)Von(e | x60 | = ey|k) (21.44) The summation over the Greek indices extends over the complete set of one-particle states. Equations (21.44) are known as the Hartree-Fock equations. From them we immediately infer that (k| Ho |k) + Ste V |kt) — (kt| V |tk)] = e, (21.45) = Exercise 21.9. Verify that Hyp, as defined by (21.44) is Hermitian. The occupied states |t) in equations (21.44) and (21.45) are not at our discretion; they must be chosen from among the eigenkets of (21.44) in a manner which will minimize the expectation value of the Hamiltonian. Frequently, the best choice corresponds to the use of those eigenkets which538 Chapter 21 Applications of Second Quantization belong to the n lowest eigenvalues ¢,, although, perhaps contrary to expec- tations, the variationally minimal value of (#) is not just the sum of the Hartree-Fock one-particle energies ¢,. Rather, the Hartree-Fock approxi- mation E, to the ground state energy is Ey = (2) = (Yo HV) = S (el Hy) +S kl V tk) — Cl Vk) ik=1 = ite + (kl Ho |k)] =30 - 2S cand V lik) — ik| V |ki)] (21.46) The state a'a,¥, (with k and j labeling occupied and unoccupied single- particle states, respectively), is orthogonal to the approximate ground state ‘f, and may thus be regarded as an approximation to an excited state of the system. In this state we have (2) = (ala,¥,, #ala,¥,) = Ey + ©; — & — (ikl V Lik) + Gikl V Iki) (21.47) If the last two terms can be neglected, e; — &, represents an excitation energy of the system. Exercise 21.10. Verify equation (21.47). Exercise 21.11. Prove that the expectation value of # in the “ionized” state a,¥, is (#) = E,—& (Koopmans? theorem) (21.48) The practical task of solving the Hartree-Fock equations is far from being straightforward. The equations have the appearance of a common eigenvalue problem, but the matrix elements of the interaction V, which enter the construction of the effective one-particle Hamiltonian Hy,y, cannot be computed without a foreknowledge of the appropriate n eigensolutions |t) of equations (21.44), The coupled equations (21.44) are thus thoroughly nonlinear and require an iteration technique for their solution. One starts out by guessing a set of occupied one-particle states |t); using these, one calculates the matrix elements of V; and one then solves the Hartree-Fock equations (21.44). If the initial guess was fortuitously good, n of the eigen- solutions of (21.44) will be similar to the initially chosen kets. If, as is more likely, the eigensolutions of the Hartree-Fock equations do not reproduce the starting kets, the eigensolutions corresponding to the lowest n eigen- values ¢, are used to recalculate the matrix elements of V, and this procedure is repeated until a self-consistent set of solutions is obtained. Sufficiently§5. The Hartree-Fock Method 539 good initial guesses of the one-particle states are usually available, so that fairly rapid convergence of the iteration process is the rule rather than the exception in actual practice. The Hartree-Fock equations can, in the representation which diagonalizes V, be rewritten as & Ul He i2MB| + See] PrvelB | 964 | = (t | B)Vapla| XB | k)] = exe | k) (21.49) As an application of these equations, we consider an atom with a nuclear charge Ze and with n electrons. Then form We choose the coordinate representation with spin as the basis |x) and |), and we denote the Hartree-Fock eigenfunctions as (ra |k) = y4(ro) The Hartree-Fock equations (21.49) are then immediately transcribed into the form 1 Ir—r'l — Evy gro) — y(ro) + tS Ef v0) AF wileo wit) de mu r St 1 r-r'| ez = frre pdro)yr’o’) dr’ = ey,(r0) (21.50) These coupled nonlinear differential-integral equations constitute the most familiar realization of the Hartree-Fock theory. If the last sum on the left- hand side, due to the exchange matrix elements of the interaction, is neglected and the term corresponding to t= k in the sum over the direct matrix elements is omitted, the equations (21.50) reduce to the simpler Hartree equations, which before the advent of fast computers were of greater practical interest than the more accurate Hartree-Fock equations. Exercise 21.12. Show that the configuration space wave function correspond- ing to the individual particle state (21.37) can be expressed as the Slater540 Chapter 21 Applications of Second Quantization determinant Yah) Y(Fa62) °° YATnon) 1 | P21) pelF202) °° * PalFnon) PHO InFn) = =] ° : Diet vn! Pal) Pn(F2F2) *** Pa(EnFn) 6. Pairing Interactions. To illustrate further the utility of the operator formalism in the treatment of many-body problems, we now consider a system of identical fermions whose one-particle energy eigenstates may be grouped in pairs by use of suitably chosen quantum numbers and which are subject to a peculiar pairing interaction rather than the most general possible two-particle interaction. In the present context, a pair of fermions is defined as two particles occupying two paired one-particle states. These are conveniently labeled by the quantum numbers m and —m. An interaction between the members of a pair of the form V =$% aia 4a_mn(gs —4| V Im, —m) (21.51) om is said to be a pairing interaction. For example, the effective Hamiltonian for electrons in a solid appears to contain a term like (21.51), provided that an electron (or Cooper) pair is defined by two electrons which have equal energies but opposite linear momenta, p and —p, and opposite spins. In complex nuclei, effective pairing occurs between nucleons in the same subshell but with opposite values of the component of angular momentum along the nuclear symmetry axis. Generally, in the absence of the interaction, the one-particle states m and —m, which constitute a pair, are degenerate in energy, denoted as e,,’ = &_m', and we shall assume this to be the case. Our assumptions make it possible and convenient to restrict all summations over the one-particle quantum numbers to positive values only and to write the Hamiltonian of the system of fermions as FH =D En! Andy + Andi) +S ay a gm m>0 am>o x [(q, —a| V |m, —m) — (4, —q| V |—m, m)] (21.52) Of primary interest here is a method for dealing with this system which is capable of producing nonperturbative solutions of the eigenvalue problem for #. The particular states to be considered are superpositions with appreciable amplitudes for a large number of pair configurations and differ§6 Pairing Interactions 541 significantly from perturbative solutions, im which one unperturbed con- figuration usually dominates. The so-called BCS states,* which are funda- mental to an understanding of superconductivity in solids, are such collective states produced by pairing; in nuclei, a pairing interaction is responsible for a conspicuous gap in the energy spectrum of the system near its ground state. The basic ideas of the theory are preserved, but the mathematics is much simplified if we assume that the pairing interaction in (21.52) is either equal to a constant —G or zero, depending on whether q and m belong to a specified subset S of the quantum numbers or not. In the case of super- conductivity, for instance, the electron pairing force is effective only for electrons near the Fermi energy. We thus assume that the Hamiltonian has the form HE => eg (Adin + A Om) — GY ah a gan Ay (21.53) m>o ames As usual in many-body theories, we now seek a transformation to a new set of creation and annihilation operators that will make the Hamiltonian correspond as nearly as possible to a system of noninteracting particles or quasiparticles. Here this objective is approached by defining for each value of m the new annihilation operators = peat Sm = Umm — Vm m (21.54) Bim = Mnd—m + Pn Uy, and v,, are numbers which we choose to be real. If we further require the normalization Up? + Vp? = (21.55) but otherwise leave u,, and v,, temporarily unspecified, the fermion anti- commutation relations for a‘, and a,, can be easily seen to imply that = alal + afef, = BmBa + BoBbm = BrBy + ByBn, = 0 Amity + Aittm = BBY + BBm = Sumi (21.56) Og Oe Exercise 21.13. Verify the anticommutation relations (21.56). The canonical transformation defined by equations (21.54) is quite different in kind from a transformation like (20.12) and (20.13) of the one- particle basis, as applied to spin } bosons in discussing the dynamics of ® The symbols BCS stand for Bardeen, Cooper, and Schrieffer, who recognized the role of the pairing interaction in producing the ground state of a superconductor.542 Chapter 21 Applications of Second Quantization magnons and to electrons in the Hartree-Fock theory. In the present trans- formation the new creation operators are linear combinations of the old creation and annihilation operators, and the number of particles is not conserved. We shall refer to the fermions which the operators a, and ft, create as quasiparticles associated with the given pairing interaction.” From (21.54) and (21.55) we derive the inverse transformation equations: Ay = Undn + PmBry ae t 1m = UmBim — Omen (21.57) The “vacuum” state Y’, with no quasiparticles, which is defined by meg = Bm¥y = 0 (21.58) is obviously not identical with the vacuum state Y of the particles. It is easy to see that the two vacuum states are related by the equations Fo = Gn + Pntnd mE, ¥° = TT Wn + OmBren eo - . (21.59) Exercise 21.14. Prove the relations between the vacuum state ‘¥) and the no-quasiparticle state 'Y’. The operator that measures the number of particles in the two paired one-particle states m, —m can easily be transformed into the quasiparticle representation. With normal ordering of the operators, ham + Ab Aon = Wa? + Un? = Vantin + BinBrn) + 2th a(thyBrn + Brit) (21.60) In our effort to transform the Hamiltonian (21.53) into the quasiparticle representation, we also require the expansion of the pairing interaction in terms of the quasiparticle operators: aa (A—mlbm = (atty + VeBe(4aBy — Pate) (mB m — Pm) (mem + Pith) If we arrange the expansion in normal product form and neglect all terms 71n honor of N. N. Bogoliubov, who invented the transformation described in this section, this particular species of quasiparticle was christened bogolon in a lighthearted introduction to many-body theories, R. D. Mattuck, A Guide to Feynman Diagrams in the Many-Body Problem, McGraw-Hill Book Company, New York, 1967. For a more advanced treatment of many-body problems, see A. A. Abrikosov, L. P. Gorkov, and I. E. Dzyalo- shinski, Methods of Quantum Field Theory in Statistical Physics, translated by R. A. Silverman, Prentice-Hall, Inc., Englewood Cliffs, 1963.§6 Pairing Interactions 543 which contain more than two quasiparticle operators, we arrive eventually at the result: tat a 4 © 400-04 mbm = DWV + Pm! + ames ames: mes: +3 [at — ra0a! — Bhan tah (ttn + BB) +3 [unt — PZ abe + 2a? |B a) The quasiparticle Hamiltonian %,,, which we obtain in this approxima- tion does not commute with N, the operator corresponding to the total number of particles. In seeking the eigenstates of the approximate Hamil- tonian it is, therefore, not possible to work in a subspace of Hilbert space corresponding to a fixed number of particles. When the full Hilbert space is available, the lowest energy state is the particle vacuum ). Since we are not interested in this trivial solution, it is necessary to impose on the varia- tional problem of #,,, the constraint (Ny =(¥,NP) =n (21.61) which insures that at least the mean value of N equals the number of fermions present in the system. We shall take this constraint into account by solving the eigenvalue problem for the operator # — AN and determining the Lagrangian multiplier 2 subsequently from condition (21.61). In the approximation described above which neglects all interactions between quasiparticles, we obtain 2 ayy — AN = 2 (en — Mog? — o( z a) +GS dg! mes mes +L Cm — Dm? = Pn Vn + BrnBrm) + 2G Suge S Undn( tin + BB) as © mes $23 Cm — DitnPn(tmBe + Bm) = GS ure S (un? — Pn VtnBrn + Bon) (21.62) as * “mes where ifm¢S and (21.63) fn = Eni —Gdyt ifmeS The eigenvalue problem posed by the operator (21.62) can be solved exactly and simply by a suitable disposition with regard to the choice of values for544 Chapter 21 Applications of Second Quantization the as yet undetermined coefficients u,, and v,,. These are selected in such a manner that the terms containing «'f* and Ba vanish identically: Umm =O if m€¢S (21.64) ug? — vg? = 2entnlem =D if mes (21.65) GY uve as It is convenient to introduce the abbreviation A=GY urn (21.66) *n From equations (21.55) and (21.65) we obtain® uy Vm = —————, meS (21.67) VEn — OP + and (21.68) (21.69) taf (1-4) 21.10) Pm = 5 En (21.70) where Em = lem — Al ifm¢S (21.71) and En =V(ém — a? + Mt ifmeS (21.72) With this notation the effective Hamiltonian can now be written as ; ae Hose — IN = 2E (Em — Dnt — T+ GE Out +S Enltntin + Bob) ™ G mes m (21.73) The quantity A is seen to be determined from equations (21.66) and (21.67) by the condition Grea eee 2 mes Jie, — a+ A? 8 If we replace um, by Ym and dy, by —Z, a second solution is obtained, but this does not correspond to a minimum df the energy. (21.74)§6 Pairing Interactions 545 A trivial solution of these equations is given by A = 0 and w=1, M0 if ee, >Ay uw =0, we =1 if em 49, and the annihilation of a particle, or the creation of a hole, if e, < Ag. In either case, an amount |e, — A | is added to the energy Ey. The quasiparticle approximation defined by A = 0 is thus equivalent to an ordinary perturbation procedure based on the noninteracting particles. The states obtained in this manner are called the normal states of the fermion system. When G is different from zero and positive, the equations for u,, and v,, admit in addition of a qualitatively different solution, with A 4 0 deter- mined by the condition (21.74), or G 1 2 Wes Jie sae eae (21.78) p+? The “chemical potential” 4, which is the analog of the Fermi energy, is defined by equation (21.61) which, for the new quasiparticle vacuum state, reduces to 2D0,2=n (21.79)546 Chapter 21 Applications of Second Quantization or x (1 - a ’) =n (21.80) ™ ‘m It should be noted that equation (21.78) has a solution only if G 1 2 me [em — Al >1 (21.81) The lowest energy state for A ¥ 0 is the state with no quasiparticles defined by (21.69) and (21.70). Great interest attaches to this particular state, denoted by Vues = [1 (in + ema’ mE (21.82) 7 if its energy expectation value Egg lies below E,, the energy of the lowest normal state. We find from (21.73) and (21.79), 2 Enos = 23 entnt —* +6 > ont (21.83) m G mes In applications to extended systems the last term in this expression may usually be neglected; also, in such systems it is a good approximation to make the identification 2 ~ A). Under these assumptions the BCS state differs from the lowest normal state by x AE = Egos — Ey = 2 2, (Em — Ao)Pm> + 2 2, Co = mn) an — e At first sight it may be confusing that in the description given in this section both the state Vy of (21.75) and the state '¥ pg of (21.82) are no-quasiparticle states, or quasiparticle vacuum states. This paradox is resolved when it is noted that the quasiparticles defined by A = 0 are quite different from those defined by A #0. Exercise 21.15. Show that the solution (21.69), and (21.70) is recovered if the expression (21.83) is minimized with respect to u,, and v,,, subject to the constraint u,,? + v,,? = 1. The physical significance of the coefficients u,, and v,, is readily appreci- ated by considering the BCS ground state (21.82). The quantity »,,” measures the probability that a pair of particles occupies the one-particle levels m and —m. As Figure 21.1 shows, v,,” drops from the value 1, indicat- ing certain occupancy of levels deep within the Fermi sea, to zero, high above the Fermi surface. The transition takes place in an energy interval of the§6 Pairing Interactions $47 iG=0 L ; Figure 21.1 fe order of A, if the pairing interaction region S extends over a broad band of one-particle energy levels surrounding the Fermi level. As equation (21.73) shows, the excited states are characterized by the creation of one or more quasiparticles, and the energy contributed by each quasiparticle is E,,. The minimum quasiparticle energy, corresponding to Em = Ay, is A, which is thus the energy gap between the quasiparticle vacuum and states with one quasiparticle. In order to evaluate the energy difference between the BCS and the normal state for a case of practical interest, we assume that the energy range S, over which pairing interactions are effective, extends symmetrically about the Fermi energy 4, so that G #0 if A, + hw > &m > dy — hw, and G = 0 otherwise. If we further assume the single particle energy levels to be dense enough and the range 2/c to be small enough compared to Ao so that the sums can be replaced by integrals with a constant density of energy levels, we have Tigo € Ae AE= 22) | a(t a 7 )® as ° Ve + AY G if fy ay earceiny Ho _ A? = p(A)(hwy [1 i + ()] + p(A,) A? arcsinh me (21.84) joo] where p(A,) is the density of energy levels at the Fermi surface, counting each degenerate level of pair states only once. On the other hand, A is determined from equation (21.78) in the same approximation as LaG{ eet adae otig *), Vera a am or ho ee ne aresinl A way548 Chapter 21 Applications of Second Quantization or 1 PAG Substituting this result into (21.84), we find that the last two terms cancel and that the remainder reduces to A = he/sinh (21.85) 2p(Ag)(heo)” AE=— ee (21.86) The analysis carried out here is tailored to an undertanding of super- conductivity.® The Hamiltonian then describes the electrons in the solid and the attractive (G > 0) pairing interaction between electrons of opposite momenta and opposite spins, resulting from the basic electron-phonon interaction through the elimination of virtual phonons. The range over which the pairing interaction is effective is defined by hw ~ kO (Debye temperature). The strength G of the interaction is inversely proportional to the volume of the solid; hence, p(A,)G is independent of the volume. The energy gap (21.85) is therefore also independent of the volume. Since the interaction is partly compensated by the Coulomb repulsion of the electrons, G is weak, so that p(A))G | and A 2heve Mole showing that the gap is much smaller than fiw. The energy gap and the distinctive, isolated BCS ground state, which the gap separates from all other states of the electron system, provide a basis for an explanation of many features of superconductivity. Exercise 21.16. Discuss the limit of strong pairing interactions. Exercise 21.17. Evaluate the variance of the total number of particles for the BCS ground state in terms of u,, and v,,. 7. Elements of Quantum Statistics. The operator formalism of Chapter 20 is ideally suited for treating large systems of identical particles in thermal equilibrium. If f denotes the average density or statistical operator for a grand canonical ensemble with fixed values for the averages of # and N, statistical thermodynamics requires that the entropy, defined as S = —k trace (f log p) (21.87) ® Many important papers are included as reprints in D. Pines, The Many-Body Problem, W. A. Benjamin, New York, 1961.§7 Elements of Quantum Statistics 549 be made a maximum subject to the constraints (N) = trace (pN) =n, (#) =trace(p#)=E, trace p= 1 in generalization of the formulas of Section 13.4. Using the Lagrangian multipliers « and 8, and the variational principle O(S — ka(N) — kp(#)) =0 it is easy to derive the central formula p= en 0/Z (21.88) where Z = trace e*N-0" (21.89) is the grand canonical partition function. For a system of noninteracting identical particles with one-particle energies ¢;, known as a perfect gas in thermodynamic parlance, # = Leala; = YeN; (21.90) The quantum statistical average of any physical quantity Q may be computed by application of the formula (Q) = trace pQ (21.91) We apply this relation to the evaluation of the average octupation numbers (A) = (aja) = trace (e-**** ala)/Z Using equations (20.27), (20.28), and (20.29), and the identity (8.106), we find that aN px ot trace (€ (a+ Bei ala) =e > trace (e*8-P* a,at) (21.92) Exercise 21.18. Verify the equation (21.92). If the commutation relations for bosons or fermions are used, we obtain (upper sign for bosons, lower sign for fermions) trace (e~*4-P gla.) = e~+#°0 trace [et "(1 F ala,)] and hence, (RN) = (aja) = (#8 F 1) (21.93) which is the familiar formula for the distributions of particles in Bose-Einstein and Fermi-Dirac statistics, respectively. The derivation given here is intended to exhibit as plainly as possible the connection between the commutation or anticommutation relations for the creation and annihilation operators and550 Chapter 21 Applications of Second Quantization —1 or +1 which characteristically appears in the denominator of the distribution law. Exercise 21.19. Use a similar technique to show that the square of the fractional deviation from the mean occupation number is i= AD) _ WA = BY _ atte 1 wy ) Problems 1. Show that if V(r) is a two-particle interaction which depends only on the distance r between the particles, the matrix element of the interaction in the k-represen- tation may be written as 1 (kskyl V kaka) = 6(ky + ke — ky — ky) aay Morne where /iq is the momentum transfer fi(k, — k,). Work out the matrix elements for the screened Coulomb potential Vye~*"/(ar), and construct the corresponding two-particle interaction operator Y for iden- tical particles in terms of the creation and annihilation operators in k-space. Exhibit the direct ((k,ks| V |kyk,)) and exchange ((kaky| V k,k,)) terms in the interaction. 2. Consider a systém of identical bosons with only two one-particle basis states, al) s¥ and at, .‘¥), Define the Hermitian operators x, pp, y, py by the relations 1 1 Ap = ae (« + #2), aia = VR (« +i if) where c is an arbitrary real constant, and derive the commutation relations for these Hermitian operators. Express the angular momentum operator (21.6) in terms of these “coordinates” and “momenta,” and also evaluate ¥*. Relate J? to the square of the Hamiltonian of an isotropic two-dimensional harmonic oscillator by making the identification c = Vu, and show the connection between the eigenvalues of these operators. 3. Using the fermion creation operators a;1,, appropriate to particles with angular momentum j, form the closed-shell state in which all one-particle states m = —j to j are occupied. Prove that the closed shell has zero total angular momentum. Ifa fermion with magnetic quantum number m is missing from a closed shell of particles with angular momentum j, show that the hole state may be treated like a one-particle state with magnetic quantum number —m and an effective creation operator (—1)*™a;», in the coupling of angular momenta. 4. Consider the unperturbed states at... °° * @im, °° * @m, 2" of n electrons, each occupying one of n degenerate orthogonal orbitals labeled by the quantum number k, and with m, = 44 denoting the spin quantum number associatedProblems 551 with the orbital k. Show that in the space of the 2” unperturbed states a spin- independent two-body interaction may, in first-order perturbation theory, be replaced by the effective Hamiltonian 1 Hose = — FD (ell V UIk)S, “Sy + const. a where S; is the localized spin operator Se = 5 LY Ahantemy imal o lone’) mum Nia 5. Consider a system of identical fermions with 2p degenerate one-particle states at, ¥ (kK =1,2,...,p) and with a pairing interaction D2 H = -G > > datayay, ike Define the fermion quasiparticles created by the operators Ble = Olen = Re and: show that # can be expressed simply in terms of the quasispin operators Sp =} D bh asborpaale'l elo”) where o’, o” are the eigenvalues +1 of c,. Show that the eigenvalues of the z component of each quasispin are }, —}, and zero, depending on whether the states k, —k are occupied by a pair of particles, vacant, or occupied by one particle (“‘broken pair”). Show that the eigenvalues of the square of the total quasispin operator > S, are expressible in terms of the total quasispin quantum number S = }(p — v), where v (the seniority) is the number of broken pairs. Derive the formula for the energy spectrum of #: Bp = —S(—0)p —n 0 +2) if n is the total number of particles in the state. Calculate the ground state (v = 0) energy approximately by the Bogoliubov quasiparticle method (Section 21.6), and compare the results for p > 1. 6. Apply the Hartree-Fock method to a system of two “electrons” which are attracted to the coordinate origin by an isotropic harmonic oscillator potential V(r) = 4uo*r? and which interact with each other through a potential c(r’ — r”)®, Solve the Hartree-Fock equations for the ground state and compare with the exact result and with first-order perturbation theory. (See Problem 5 in Chapter 15.) 10M. Moshinsky, Am. J. Phys., 36, 52 (1968).552 Chapter 21 Applications of Second Quantization 7. For a Fermi gas of free particles with Fermi momentum p,, calculate the ground state expectation value of the pair density operator LH OWL EY Mr) in coordinate space and show that there is a repulsive correlation which would be absent if the particles were not identical. Show that there is no spatial correlation between fermions of opposite spin. 8. In a superconductor the pairs are defined in terms of free electron momentum eigenstates with momenta p and —p and opposite z components of the spins. If the operator which creates an electron with momentum p (—p) and spin up (spin down) is denoted by a},/.(at 1/2), the pairing interaction takes the form GY ah yj20th yr 1/2» 1241/2 Do Show that the expectation value of the pair density operator LOOM OY40) + LOYUCOYL OC) for electrons with opposite spins in coordinate space exhibits spatial correlations which decrease with increasing distance between the two electrons. (Do not attempt a complete evaluation of the integrals in momentum space.)!! 9. Calculate in first order the energies of the 1S, 3P, and 'D states arising from the atomic configuration p®. Use the expansion & ar TO Zymed YZ (DEO MEE for the interaction energy between the electrons, and show that the term energies may be expressed as ECS) = Ey + (79) + 38472) EGP) = Ey + (v0) — as(v2) ECD) = Ey + (%) + 26(72) where (7,) is the radial integral (re) = ff le RO YPLRO Pre”? dr’ de” (Note. This is a somewhat difficult problem.) 1 For details on estimating the integrals, see J. Bardeen, L. N, Cooper, and J. R. Schrieffer, Phys. Rev., 108, 1189 (1956), Appendix D.