You are on page 1of 8

February 1972

Introducing Electron Creation and Annihilation Operators in Solid State Physics



M. I. DARBY

Department of Pure and Applied Physics University of Salford

Salford 5, Lancashire, England

(Received 29 June 1971; revised 31 August 1971)

A brief and simple introduction to the theory of many (N) electron systems that is suitable for undergraduates is given. By considering only three electrons (N = 3) it is possible to write out all the results explicitly. The manipulations of both determinantal wavefunctions and wavefunctions in the number representation are considered, and the electron annihilation and creation operators are defined. It is then shown how the Hamiltonian for N electrons is written in second quantized form.

INTRODUCTION

As knowledge of physics advances, it becomes necessary at some stage to introduce the new concepts to undergraduates in order that they at least recognize the basic ideas and terms used in the current scientific literature. In view of this, the time now seems appropriate to begin to extend the material content taught concerning the theory of electrons in solids. The single electron theory, which is always included in undergraduate courses, assumes that the electrons are independent, and only a one-electron Sehrodinger equation is necessary to describe their properties. In fact the motions of electrons are quite strongly correlated through the Coulomb interactions between them, and the theory of electrons in solids is strictly a many body problem. Over the last decade it has increasingly been treated as such.

It is found that the many body effects are important in determining properties such as the cohesive energy and dielectric constant in solids, and they predict plasma oscillations that have no counterpart in single particle theory. On the

other hand, much of many body theory has shown that the single particle theory is reasonably valid. One reason for this is that a given electron is well screened from a neighboring electron by all the other electrons, so that the electrons, plus screening clouds, which are usually known as "quasiparticles," interact only weakly. It also turns out that the low lying excited states of the electronic system, which are important in determining its properties, can be described in terms of excitations of one or more quasiparticle from the ground state. These are often referred to as "elementary excitations."

In general, the many body problem has been thought to be too difficult to be taught as part of an undergraduate course, with the result that many graduate experimental solid state physicists, who have not had further training in theoretical physics, find that even the basic concepts and terminology are a barrier in the further understanding of their subject. Frequently the first stumbling block is a lack of knowledge about creation and annihilation operators. Invariably, or so it seems, a model calculation of a given property of a solid begins by writing down the Hamiltonian in terms of these operators; usually it is said to be written in second quantized form. Consequently it is suggested, as a beginning, that the concepts of electron creation and annihilation operators be introduced in undergraduate courses. Unfortunately the more advanced texts! often consider the basic properties of these operators too briefly to be suitable for study at the required level. In this article, therefore, a simplified introduction, which is much more detailed than is found in those texts, is presented. It is hoped that it will also be of interest to some graduate students.

For brevity, only the electron properties of a solid are considered here, and the aim is restricted to explaining, in detail, how the Hamiltonian is written in terms of annihilation and creation operators. The main advantage of using this second quantized form for the Hamiltonian is that it allows the many body wavefunction to be written in a form simpler than the Hartree-Fock

AJP Volume 40 / 281

M. I. Darby

determinant, and consequently matrix elements are more easily evaluated. Further, the low lying excited states of the system are readily described in terms of annihilating one electron in the ground state and creating another of higher energy. The formalism is particularly suitable for describing the quasiparticles mentioned above.

In what follows the determinantal wavefunction is described, the electron creation and annihilation operators are defined, their simple manipulations are given, and it is shown why and how. the Hamiltonian of an interacting electron system is written in terms of them. To simplify the theory as much as possible, only three electrons are considered, and their spins are neglected. The results are easily generalized to many electrons.

The theory of two electrons, as applied for example to the helium atom or the hydrogen molecule, is included in most undergraduate courses. It is possible with so few particles to go a long way towards solving the Schrodinger equation exactly, and it would be superfluous to use many body theory. However, some of the basic results for two electrons are important, and we shall briefly review them here.

The theory for three electrons is so much more difficult than for two that it can be taken as the simplest example of a many body problem. It has, however, the great advantage over the N-particle problem in that the amount of labor required to write everything down is not prohibitive. For this reason the three body problem is useful in presenting the basic ideas, and the results are easily generalized to the case of N particles.

It is common in most introductions to many body theory to assume a prior knowledge of determinantal wavefunctions and matrix elements of operators between them, but frequently this assumption is unjustified. Indeed, many standard introductory texts to quantum theory, e.g., Schiff," do not discuss the manipulations of determinantal wavefunctions at all. Yet familiarity with such manipulations is essential because in using electron creation and annihilation operators we seek to represent determinantal wavefunctions in a different way. Thus the determinantal wavefunctions for three electrons are discussed below before defining the creation and annihilation operators.

282 / February 1972

TWO ELECTRONS

The quantum theory of two spinless electrons serves to introduce many of the basic concepts (and approximations) of many body theory, but since it is included in most texts, it is only described here very briefly.

The energy of two electrons moving in a common potential V(r) is determined by a Hamiltonian which consists of contributions from the electrons moving independently, plus a Coulomb interaction term between them:

H = - (fi,2/2m) v?+ V(rl) - (r"z;2m) '\7z2 +V(r2)+(e2/1 rl-r21), (1)

where subscripts 1 and 2 refer to the two electrons. The system is described by a total wavefunction 'I! (r., r2) that satisfies the Sehrodinger equation Hw=E'I!.

A basic approximation in the theory is to assume that the electrons move independently, which means that they can be described by single particle wavefunctions, if;(rl) and cp(r2) say. The simplest two particle wavefunction consistent with the statistical interpretation of 'I!*'I! is the product if; (rr) cp (r-), and in many electron theory a simple product like this is known as a Hartree wavefunction. Unfortunately the simple product does not satisfy the Pauli exclusion principle since both electrons can be in the same state, cp say, and clearly 'I!=cp(rl)cp(r2) does not vanish, as it should. In fact it is an essential property of electronic wavefunctions that they are antisymmetric with respect to interchange of two electrons, i.e., we require

This property can be satisfied by choosing the following antisymmetrized product of the single particle functions if; and cp:

which may be written as the determinant

The exclusion principle follows immediately because 'l' vanishes if If; is the same as cpo The factor 1/V2 is a normalization constant, it being assumed that cp and If; are orthonormal. Antisymmetrized products like 'l'(rl' r2) are known as Hariree-Fock waveJunctions, or often as Slater determinants, and we are concerned exclusively with them here.

The expectation value for the energy is

E= J J'l'*(rr, r2)H'l'(rl, r2)drIdr2

= J J[lf;*(rl)cp*(r2) -cp* (rl)lf;* (r2) ] XH[lf;(rl)cp(r2) -cp(rl)lf;(r2) ]drIdr2

=I+J,

where

1= J flf;*(r) cp*(r')Hlf;(r) cp(r')drdr'

and

J = J flf;*(r)cp*(r')Hlf;(r')cp(r)drdr'.

I is a Coulomb-type integral that involves the charge densities If;* (r) If; (r) and cp* (r) cp (r), while J has the form of an exchange integral that is not present within the Hartree approximation and is a consequence of the antisymmetry of 'l'. The single particle terms of H, such as -jli2V'N2m+ V (r.), do not contribute to J because the integral

flf;* (r-) cp(r2) dr2J cp* (r.)

X [ - (jli2/2m) V'12+ V(rl) ]If;(rl)drl

vanishes due to the orthogonality of If;(r2) and cp (r-). For finite exchange there must be an interaction between the electrons, and in practice, as here, this is usually the Coulomb interaction e2/1 rl-r21·

THE THREE ELECTRON PROBLEM

The Hamiltonian for three electrons moving in a potential V (r) is a simple extension of Eq. (1), but it is convenient to write it as the sum of single and two-particle operators in the form



H= 'LJ(ri)+! 'L v(ri-rJ),

i=l

i,i=1

where

Electron Creation and Annihilation Operators

is the kinetic plus potential energy of the ith electron and

is the Coulomb interaction between electrons i and i Each electron is assumed independent of the others with its own normalized wavefunction. If these single particle functions are If;(rl), cp(r2), and nr.) , then an antisymmetric total wavefunction 'l' (r., r2, r.) can be constructed as the Slater determinant:

.\Crl) s(r2) s(r.)

= (3!)-1/2{lf;(rl) [cp(r2)'\cr.) -cp(r.)s(r2)] -If;(r2) [cp(rl)'\cr.) -cp(r.)'\crl)]

+If;(r.) [cp(rl)S(r2) -cp(r2)t(rl) ]}.

If two columns of the determinant are interchanged its value is multiplied by -1, so that antisymmetry is ensured, while 'l' vanishes if two rows are the same, so that the exclusion principle is satisfied.

It is straightforward, and instructive, to write out the matrix elements of "L,;](ri) between 'l'(rl' r2, r.) defined above and

The results are identical in form for any single particle operator. The contribution from j'(r.) is

(2)

(3!) J'l'*J(rl)'l"drIdr2dr.

= Jcp* (rs) q/ (r-) dr2Js* (r.) r' (rs) ds,

X flf;* (rl)J(rl)lf;' (rl) drl- J cp* (r2) s' (r2) dr2 X Js* (rs) cp' (rs) dr.flf;* (rl)J(rl) If;' (r.) dri

+ Js* (rs) t' (r2) dr2J cp* (rs) cp' (rs) dr.

X flf;*(rl)J(rl)lf;' (rl)drl+'"

=6FI, say.

AJP Volume 40 / 283

M. I. Darby

It is clear that:

(1) If all of v/, cf>', and (differ from if;, cf>, and r, then all terms vanish due to the orthogonality of the single particle functions.

(2) Even if two of if;', cf>', and r' differ from if;, cf>, and r. all terms still vanish.

(3) If only one of if;', cf>', and t' differ from if;, cf>, and r, then two terms do not vanish.

Consider the third case and suppose that cf>'==cf>, r'==r, but if;'~if;, then the first and third integrals do not vanish and

Similarly

with the result that

3

L J'fI*f(ri)'fI'dr1dr2dra= Jif;*(r)f(r)if;'(r)dr.

i=l

It is clear that in evaluating the diagonal matrix elements of L;f(ri) fewer of the integrals involved vanish and it is readily found that

i=1

i=l

where the notation is changed to if;1 == if;, if;2 = cf>, if;a==r, so that the result can be generalized to N particles by increasing the terms in the summation. It is often convenient to write the matrix elements in terms of the Dirac bra and ket vectors, e.g.,

L ('fI I fi I 'fI)= L (i I f Ii>. (3)

i i

The quantity (i I f I i) is just the expectation energy of the ith electron assuming it is completely independent of the others, and the diagonal element therefore gives the sum of these energies.

The matrix elements of Li,j v(ri-rj) are much more tedious to evaluate, but it can be shown that individual matrix elements like J'fI*v'fl'dr1dr2dra vanish if more than two single particle wavefunctions in 'fI and 'fI' are different. Suppose that r'==r but if;'~if; and cf>'~cf>, then by writing out all

284 / February 1972

the terms, as previously, one finds

t L J'I!*v(ri-rj)'fI'dr1dr2dra i,i

= J Jif;* (r) cf>* (r')v(r-r')if;' (r) cf>' (r') drdr'

- J Jif;* (r)cf>*(r')v(r-r')if;' (r')cf>' (r)drdr' == (ij I v I ij)- (ij I v I ji),

in the Dirac notation. Here if;i=if; and if;j=cf>, so that the results can be immediately generalized to more particles. The diagonal matrix elements are

= t L I (ij I v I ij) - (ij I v I ji)}. (4)

i,i

The two terms within the summation are the Coulomb and exchange integrals between two electrons, which were discussed above. Expression (4) is therefore the sum of these energies for each pail' of electrons taken in turn, and the factor t (= 1/21) corrects for the fact that the sum as written counts each pair twice.

Finally the total average energy of a system of N electrons which have a Hamiltonian like Eq. (2) IS grven by the sum of expressions (3) and (4), i.e.,

N

E = L (i I f I i)

i=1

+t L I (ij I v I ij)- (ij I v I ji)}. (5)

i,i

ELECTRONS IN METALS

The primary interest here is to consider the theory of N electrons in a metal, and for illustrative purposes it is assumed that N = 3. In the simplest free electron model the potential V (r) is zero everywhere, and it is well known" that the appropriate single particle wavefunctions are the plane waves

where n is the crystal volume. These states are uniquely labeled by the wave vector k, and each

electron has (kinetic) energy n,2k2/2m. If there are N electrons, then at zero temperature they occupy the N lowest single .particle energy states. The highest occupied energy level is called the Fermi energy E" and there is a corresponding largest wavenumber k! defined by Ef=h2ki/2m. The probability nk of finding an electron with wavenumber k is unity for k~k" and zero otherWIse.

With N ~102S it is clearly impossible to write down the NXN Slater determinant for the metal. However, it is slightly less impracticable to write down the N occupied wavefunctions or since it is assumed that the states only differ in the wavenumber k, to list the occupation numbers nk of the occupied states. [This assumption is true even when V (r) is nonzero, so that what follows is also applicable when the electrons have Bloch wavefunctions l/tk(r).J In the case N =3 with the lowest single particle states specified by wavevectors k, p, and 1, the total ground state wavefunction may be written as

where the unoccupied states with n = ° are suppressed. In this form the wavefunction is said to be in the number representation; the Slater determinant is sometimes referred to as the configuration representation. It is important that the order in which the occupation numbers are listed in (6) is used consistently, because interchanging two of them changes the sign of the total wavefunction. Thus, for example, if

Y;k (r.) Y;k (r2) Y;k (fa)
Ink, np, nl)= (3!)-1/2 Y;P (r.) Y;P (r-) Y;P (rs)
Y;I (r-) Y;I (rs) Y;I (rs)
then
Y;k(rl) Y;k(f2) Y;k (rs)
Ink, nl, np)= (3!)-1/2 lh (r.) Y;I (r2) Y;1(rs)
Y;P (fl) Y;P (r2) Y;P (rs)
= (-1) Ink, np, nl). Electron Creation and Annihilation Operators

Excited states have the form Ink, np, nl = 0, ni) == Ink, np, ni), Ink, np=O, nl, nj)==1 nk, nl, nj), etc., in which one or more electrons are promoted above the three lowest single particle energy levels.

ANNIHILATION AND CREATION OPERATORS

The task of computing matrix elements involving wavefunctions in the number representation is facilitated by introducing electron annihilation and creation operators. Basically one is trying to replace a conventional operator, F say, by another, Fop, such that the matrix elements of Fop between wavefunctions I n) in the number representation are identical to those of F between the equivalent determinantal wavefunctions \(F, i.e., such that

(n I Fop I n')== (\(F I F I \(F').

It is known a posteriori that Fop can be expressed as a function of annihilation and creation operators, and it is therefore first necessary to define these operators and discuss some of their properties before considering an example.

The annihilation operator for an electron, Cp, is defined such that when it operates on a wavefunction it annihilates one electron in state Y;P' e.g.,

The order of the occupation numbers is important. For example,

Further, the result of operating with C; upon a state in which Y;P is unoccupied gives zero since it is impossible to annihilate an electron in a state which is not present, e.g.,

The creation operator cpt is defined such that when it operates on a wavefunetion it creates an electron in state Y;P' e.g.,

AJP Volume 40 / 285

M. I. Darby

and the order of the n's should be noted. By inter- Clearly changing the occupation numbers one also has

If state ,pp is already occupied, then since there can only be one electron per state, the result of operating with cpt gives zero, e.g.,

The results for three particles given above are generalized to N particles in the definitions

Cpl"'nk,np,nl, ... )=(-l)ml "'nk,nl, ... ), cpt 1 .. 'nk, nl, ••• )= (_l)m I·· 'nk, np, nl"'),

where m is the number of occupied states preceding np in the wavefunction.

We now discuss some simple properties of C p and cpt and in particular define the number operator. Consider the effect of first annihilating and then recreating a single particle state ,pp, i.e., of Cp tcp acting upon 1 tu; np, nl). The operator nearest the wavefunction acts first:

Then, operating with c, t,

CptlCp 1 nk, np, nl)l = -cpt 1 nk, nl) = -I np, nk, nl)

i.e.,

However, if state ,pp is originally unoccupied, the first operation gives zero, so that

Thus it is possible to determine whether the state ,pp is occupied in any many-particle wavefunction by operating upon it with Cp tcp. For this reason Cp+Cp is known as the number operator, and it is often denoted by N p:

286 / February 1972

p

In a similar way, it is straightforward to obtain the results for other pairs of creation and annihilation operators, such as CpCp t, C, tCp, etc. As a further example, consider the effects of C, tcp and CpCi t upon 1 nk, np, nl) when ir£p:

C; 1 nk, np, nl) = -I nk, nl), CitlCp 1 nk, np, nl)l = -I ru, nk, nl).

Also

= Cp 1 np, tu, nk, nl) = 1 ni, ru, nl).

Adding these results gives

It can be shown that the effects of all such operations are summarized in the anticommutation relation

(10)

The left-hand side is often written as [Ck t, CIJ+. It is easy to prove other relations such as

and

All the properties of the system of electrons that arise from the antisymmetry of its wavefunction are incorporated in the anticommutation relations (10) and (11).

THE HAMILTONIAN

(9)

As indicated at the beginning of the last section, in order to use the number representation for a many body Slater determinant to calculate matrix elements of an operator, it is necessary to obtain

the appropriate form of the operator in terms of the annihilation and creation operators. As an important example we consider the Hamiltonian for three electrons, Eq. (2), and give a plausibility argument for its equivalent form Hop. A more general and rigorous method has been given by Pike.4

Consider the single particle terms

"L 1(ri).

i=l

It has been seen that the matrix elements of this operator between Slater determinants 'I' (1/Ik, 1/Ip, 1/11) and '¥' (1/Iu, 1/1 v , 1/Iw) vanish unless only one of 1/Iu, 1/1 v , and 1/Iw differ from the corresponding unprimed single particle states. The nonvanishing terms have the form

f'¥(1/Ik, 1/Ip, 1/11) "L1(ri)'¥'(1/Ik, 1/Ip, 1/Ij)dr1dr2dra i

=(lI1Ij)·

We seek an operator, 10p say, that gives the same result when averaged between wavefunctions 'I' and '¥' expressed in the number representation, i.e., such that

The required operator is

1op=

(s 111 t)C. tCI,

as can be shown by direct substitution. Consider the term s=k, t=k:

(nk' np, nl I (k 111 k)Cz ia, Ink, np, nj)

= (k 111 k)(nk' np, nil ne, np, nj)

= 0 (by orthogonality).

In the same way it can be shown that all terms vanish except that for s=l, t=j. The nonzero term

is readily evaluated by noting that

Electron Creation and Annihilation Operators

and

Thus

(nk' np, n, I (ll 1 I j)CI 'C, Ink, np, nj)

= (ll 1 I j)(nk' np, nl Ink, np, nl)

=(lI1Ij),

which verifies the form of 1op.

The matrix elements of a two-particle term v(ri-rj) are more complicated. They vanish if more than two single particle orbitals are different in 'I' and '¥', and otherwise they have values of the form I (pll v I ij) - (pll v I ji)}. The corresponding second quantized form of v may be shown by explicit evaluation of the matrix elements to be

and the complete Hamiltonian in terms of annihilation and creation operators is therefore

N

Hop= "L (s ! f I t)C. sc,

8, t

N

+! "L (st I v I ij)C. so, sec, (13)

8, t,i,j

The expectation value of the energy for a state I ... tu, .•. nj, ••• ) is, of course, by definition of Hop, identical to that given by Eq. (5). Result (13) is a starting point for the theory of many electrons.

THE VACUUM STATE

Finally, annihilation and creation operators can also be used to give a different form for the wavefunction in the number representation that is frequently useful. Thus it is possible to construct a state with many particles by repeatedly applying creation operators Cit, C, t, ... , etc., to a state which contains a small number of particles. In the limit one can choose an initial wavefunction with no single particle orbitals occupied. Such a state is called a vacuum state, 10), and usually it is assumed to be normalized. Then, for example,

AJ P Volume 40 / 287

M. I. Darby

where it should be noted that Cit operates first, and so on. The complex conjugate of this wavefunction is

where now Ci operates first, etc.

It is an instructive example in the manipulation of the operators to show that these wavefunctions are normalized. The operators linked together are replaced in the next step by those in square brackets, and each replacement is made using the anticommutation relations (10) and (11):

..,

= (0 I CICpCkCktCptClt 10)

,--,

=(0 I CICp(I-CktCk)CptClt I 0)

= (0 I CICpCptClt I 0)

,--,

-(0 I C1CpCkt(-CptCk) Cit I 0)

= (0 I C1CpCp tCI t I 0)

+ (0 I CICpCk sc, t (-CliCk) 10) = (0 I CICpCp tCI} I 0)+0,

1 See, e.g., C. Kittel, Quanturn Theory of Solids, (Wiley, New York, 1963); J. M. Ziman, Elernents of Advanced Quanturn Theory (Cambridge U. P., Cambridge, Eng., 1969); P. W. Anderson, Concepts in Solids (W. A. Benjamin, New York, 1963); N. H. March, W. H. Young, and S. Sampanthar, The Many Body Problern in Quanturn

288 I February 1972

because G\ I 0) = 0; there is no single particle state in I 0) to annihilate. The process can be repeated by using the anticommutation relations to transpose first C p and then C I to be next to I 0), so that successively

(nk' np, nl Ink, np, nl) = (0 I ClCpCp tCI t I 0) = (0 I CIClt I 0)

= (0 I 0)

=1.

Orthogonality may be proved in the same way.

CONCLUSION

The account above shows that for a system of N electrons the Hamiltonian in second quantized form, Eq. (13), is a plausible equivalent to the conventional one, Eq. (2). While this is only a very preliminary step towards understanding many body theory (and in fact nothing has been said here of the process of second quantization), it is often the first step that is most difficult. It is hoped that this method of presentation, using explicit results for three particles, will enable the subject to be introduced more widely at the undergraduate level.

Mechanics (Cambridge U. P., Cambridge, Eng., 1967).

2 L. 1. Schiff, Quanturn Mechanics (McGraw-Hill, New York, 1968), 3rd ed,

3 See, e.g., C. Kittel, Introduction to Solid State Physics (Wiley, New York, 1956), 2nd ed.

4 E.. R. Pike, Proc. Phys. Soc. 81, 427 (1963).

You might also like