A mode expansion expresses position-space field operators in a basis of one-particle modes. It is the second-quantized version of expanding a wavefunction in an orthonormal basis, but the coefficients are creation and annihilation operators rather than complex amplitudes.
For a complete orthonormal one-particle basis {φi(x)}, the nonrelativistic annihilation field is written
ψ(x)=i∑φi(x)di,ψ†(x)=i∑φi∗(x)di†.
Here di,di† stand for either bosonic mode operators ai,ai† or fermionic mode operators ci,ci†, with the appropriate algebra. The same formulas work for both statistics; the difference is in the commutators or anticommutators of the mode operators.
The expansion is not a new postulate independent of mode occupations. It is the coordinate representation of the statement that particles or excitations occupy one-particle modes.
Let h be the one-particle Hilbert space. In the coordinate representation, an orthonormal basis is a set of wavefunctions φi(x)=⟨x∣φi⟩ satisfying
∫d3xφi∗(x)φj(x)=δij.
Completeness means
i∑φi(x)φi∗(y)=δ(3)(x−y),
understood in the distributional sense. If the basis is finite or truncated, the same sum gives the kernel of the projection onto the retained subspace, not the full delta function.
The basis may use trap eigenfunctions, periodic-box plane waves, spin-orbitals, or orthonormal wavepackets. Bound eigenstates alone are not complete if the one-particle Hamiltonian also has scattering states; its spectral expansion then requires the continuum contribution. Lattice-site orbitals generally describe a retained band subspace. The mode label i is shorthand for all quantum numbers needed to specify one complete one-particle state.
For each one-particle mode φi, there is an annihilation operator di and a creation operator di†. Acting on the vacuum,
di†∣0⟩=∣1i⟩,di∣0⟩=0.
For a normalized wavepacket f(x), the corresponding mode operator is a linear combination of basis-mode operators:
df=i∑⟨f∣φi⟩di,df†=i∑⟨φi∣f⟩di†.
In position-space notation this becomes the smeared field operator
df=∫d3xf∗(x)ψ(x),df†=∫d3xf(x)ψ†(x).
This smearing is conceptually important. The object ψ(x) is an operator-valued distribution, so sharply localized expressions should be interpreted through integrals against wavepackets, test functions, or basis modes. The operator meaning of these fields is developed in Field Operators.
Thus the expansion coefficients are operators obtained by projecting the field onto the chosen one-particle basis, exactly as ordinary wavefunction coefficients are obtained by inner products.
At coincident points its limiting value is (2K+1)/L, not an infinite delta-function peak. This projected kernel approaches the periodic delta distribution as the cutoff grows; it is not a pointwise approximation to an ordinary function. Spatial projection is distinct from a bosonic occupation cutoff, which changes the ladder commutator itself.
Suppose two orthonormal one-particle bases are related by a unitary matrix U:
∣χα⟩=i∑∣φi⟩Uiα.
In wavefunction form,
χα(x)=i∑φi(x)Uiα.
The field operator is basis-independent:
ψ(x)=i∑φi(x)di=α∑χα(x)dα.
The mode operators therefore transform as
dα=i∑Uiα∗di,dα†=i∑Uiαdi†.
Because U is unitary, the transformed operators obey the same bosonic commutation relations or fermionic anticommutation relations. The occupation labels change, but the field operator and the Fock-space state do not change.
This is an ordinary one-particle basis rotation. It should not be confused with a Bogoliubov transformation, which mixes creation and annihilation operators and may change which state is called the vacuum.
For a cubic box of volume V=L3 with periodic boundary conditions, the normalized plane-wave modes are
φk(x)=V1eik⋅x,k=L2πn,n∈Z3.
The field expansion is
ψ(x)=V1k∑eik⋅xdk,
with inverse
dk=V1∫Vd3xe−ik⋅xψ(x).
For bosons, [dk,dq†]=δkq; for fermions, {dk,dq†}=δkq.
In infinite volume, a common continuum convention is
ψ(x)=∫(2π)3/2d3keik⋅xd(k),
with
[d(k),d†(q)]=δ(3)(k−q)
for bosons, or the corresponding anticommutator for fermions. Since ∑k→V∫d3k/(2π)3, this convention relates the operators by d(k)=V/(2π)3dk in the continuum normalization limit. The discrete operator is dimensionless; d(k) has units of length3/2 and ψ(x) has units of length−3/2. Other Fourier conventions redistribute powers of 2π consistently.
When the particle has spin or another discrete internal label, the complete one-particle mode includes that label. A convenient basis for spin-1/2 particles is
∣n,s⟩=∣ϕn⟩⊗∣s⟩,s∈{↑,↓}.
The field has components
ψs(x)=n∑ϕn(x)dns,ψs†(x)=n∑ϕn∗(x)dns†.
For spinful fermions,
{ψs(x),ψt†(y)}=δstδ(3)(x−y),
and for spinful bosons the same formula holds with a commutator instead of an anticommutator.
The local number density is
n(x)=s∑ψs†(x)ψs(x),
and a spin-density component is written
Sa(x)=st∑ψs†(x)(Sa)stψt(x).
For physical spin density, use Sa=ℏσa/2 for spin one-half; using Pauli matrices without this factor instead defines a dimensionless polarization density. In a spin–orbit eigenbasis the expansion uses spinor components φi,s(x) rather than a factorized orbital and spin label.
The internal labels matter: For electrons, the Pauli principle applies to complete spin-orbitals, so two electrons with opposite spin can occupy the same spatial orbital because they occupy different one-particle modes.
Inserting the expansion into A=∫d3xd3yψ†(x)A(x,y)ψ(y) gives A=∑ij⟨φi∣A∣φj⟩di†dj. The one-body operator derivation explains why this acts as the sum of the same one-particle observable over all particles. Expanding four fields similarly produces two-body matrix elements.
Invert the mode expansion. Starting from ψ(x)=∑jφj(x)dj, prove that di=∫d3xφi∗(x)ψ(x).
Solution
Insert the expansion:
∫d3xφi∗(x)ψ(x)=j∑(∫d3xφi∗(x)φj(x))dj.
Orthonormality gives the inner integral as δij, so the sum reduces to di.
Check a basis change. If ∣χα⟩=∑i∣φi⟩Uiα and dα=∑iUiα∗di, show that ψ(x)=∑iφi(x)di is also ∑αχα(x)dα.
As a finite-lattice check, take Ns sites at rj=ja, j=0,…,Ns−1, with periodic boundary conditions. For km=2πm/(Nsa), m=0,…,Ns−1, define ckmσ=Ns−1/2∑je−ikmrjcjσ. Show explicitly that these fermionic operators satisfy {ckmσ,ckm′σ′†}=δmm′δσσ′.
Solution
Substitute both definitions:
α∑χα(x)dα=αij∑φi(x)UiαUjα∗dj.
Unitarity gives
α∑UiαUjα∗=δij.
Therefore
α∑χα(x)dα=j∑φj(x)dj=ψ(x).
For the lattice transform, use the site anticommutator to obtain
The finite geometric sum is Ns for m=m′ and zero for distinct momenta in the stated grid. The symmetric Ns−1/2 normalization therefore preserves the canonical algebra.
Spinful number density. For spin-1/2 fields, show that