NOTES ON WICK’S THEOREM IN MANY-BODY THEORY

Luca Guido Molinari

Dipartimento di Fisica, Università degli Studi di Milano,
and INFN, Sezione di Milano, Via Celoria 16, Milano, Italy

Revised: January 2016

I. INTRODUCTION II. EXAMPLES

For bosons and fermions, destruction and creation
operators of a particle in a state |ii annihilate the vac-
uum: ψi |vaci and hvac|ψi† = 0. Since observables have
null expectation value in the vacuum state, it is conve-
nient to construct them with creation operators on the
left of destruction operators (normal order).
In a many body theory one usually makes reference
to the ground state |gsi of some non-interacting or ef-
fective theory, which is filled with particles or quasi-
particles. The theory is also supplied with a basis of
canonical operators αa and αa+ , a = 1, 2, . . .
+ + +
[αa , αb ]∓ = 0, [αa , αb ]∓ = 0 [αa , αb ]∓ = δab
([a, b]∓ = ab ∓ ba) that annihilate the reference state:
+
αa |gsi = 0, hgs|αa = 0.
The Hamiltonians in the examples below share a
common property: they are all quadratic. They are di-
agonalized by a linear canonical transformation which
assures us that [A− +
i , Aj ]∓ are c-numbers.
A. Independent fermions
This example is relevant for the perturbation theory
with N interacting fermions. In the zero order descrip-
tion, the two-particle interaction is turned off and the
independent fermions are described by a Hamiltonian of
†
P
the form H = a ~ωa ca ca (a is a label for one-particle
states, ordered so that ω1 ≤ ω2 ≤ . . .). The ground
state |F i is obtained by filling the states a = 1 . . . N .
(1) The operators that annihilate |F i are:

Since they are a basis, operators ψi or ψi† , which will be

(
c†a if a ≤ N ,
here indifferently denoted as Ai , have a decomposition αa = , αa |F i = 0 (5)
ca if a > N
Ai = A−i + Ai
+
(2) (
ca if a ≤ N ,
where the first term is a combination of operators αa ,
+
αa = †
, hF |αa† = 0 (6)
+
c a if a > N
and the latter is a combination of operators αa . One
term is not the adjoint of the other: the labels − and +
αa+ creates a particle (above the Fermi level) or creates
refer to their action on the reference state |gsi:
a hole (by removing a particle in the “Fermi sea” |F i);
A−i |gsi = 0, hgs|A +
i = 0 (3) αa destroys a particle above the Fermi sea, or destroys
a hole (by adding a particle in the Fermi sea). These
Since the operators αa and αa+ are canonical, by con-
particle-hole operators form a CAR.
truction one has
Any destruction or creation operator admits a decom-
[A− i , A −
j ]∓ = 0, [A +
i , A +
j ]∓ = 0 (4) position in this basis into positive and negative parts:
while mixed brackets [A− +
i , Aj ]∓ are in general non-
X X
ψi = hi|aica + hi|aica
zero. We require them to be c-numbers1 . This is a a≤N a>N
vital assumption, which makes the Wick’s theorem hold. X X
= hi|aiαa+ + hi|aiαa = ψi+ + ψi−
In a many particle theory one encounters the prob- a≤N a>N
†
hi|ai∗ c†a = (ψi† )− + (ψi† )+
X X
lem of expanding products of several field operators into ψi = hi|ai∗ c†a +
†
normal-ordered expressions of the operators αa and αa . a≤N a>N
The general problem of bringing products of field op-
erators into a normal form was solved in 1950 by Gian with (ψ † )− = (ψ + )† and (ψ † )+ = (ψ − )† .
i i i i
Carlo Wick [1] (1909-1992). He obtained his theorem
while in Berkeley, in the effort to give a clear deriva-
tion of Feynman’s diagrammatic rules of perturbation B. Bogoliubov transformation
theory. In this example |gsi is the ground state |BCSi of the
1 since [α , α+ ] = δ , this is certainly true if the operators A± superconducting state at T = 0 (Bardeen, Cooper and
a b ∓ ab
are linear combinations of the α±
a. Schrieffer, 1957). It is filled of Cooper pairs of electrons

1
(spin singlets, with zero total momentum), III. NORMAL ORDERING AND
CONTRACTIONS
(uk + vk a†k↑ a†−k,↓ )|vaci
Y
|BCSi =
k A product of operators A± i is normally ordered if
all factors A−
i are at the right of the factors A+
j :
uk and vk are complex amplitudes, with |uk |2 +|vk |2 = 1
+ − −
for the normalization of the state. The state is not A+1 · · · Ak Ak+1 · · · An (12)
an eigenstate of the total number operator. It is an-
nihilated by the following operators (Bogoliubov and In+particular, +
a product of operators of the same type,
− −
Valatin, 1958): A 1 · · · A k or A 1 · · · A` , is normally ordered. The very
usefulness of the definition is the obvious property that
αk = uk ak,↑ − vk a−k,↓ , β−k = uk a−k,↓ + vk ak,↑ (7) the expectation value on |gsi of a normally ordered op-
† † †
erator is always zero:
−
hgs|A+
1 · · · An |gsi = 0 (13)
αk |BCSi = 0, β−k |BCSi = 0 (8)
It is clear that any product of operators A1 A2 . . . An can
Together with their adjoint operators, be written as a sum of normally ordered terms. One first
−
writes every factor as A+ n
i + Ai and gets 2 terms. In
hBCS|αk† = 0 hBCS|β−k
†
=0 (9) −
each term, the components Ai are brought to the right
by successive commutations (bosons) or anticommuta-
they satisfy the CAR rules. They are obtained by a tions (fermions). After much boring work, the desired
canonical transformation that mixes creation and de- expression will be obtained. Wick’s theorem is an effi-
struction operators for spin-momentum states, in a way cient answer to this precise problem: to write a prod-
adapted to the BCS state. uct A1 . . . An as a sum of normally ordered terms. The
Inversion gives the operators ak,σ and a†k,σ as sums theorem is an extremely useful operator identity, with
of a − term (linear combination of α and β) and a + important corollaries. To state and prove it, we need
term (linear combination of α† and β † operators). Note some technical tools.
that, although hak,σ i = 0, BCS-expectation values of The normal ordering operator brings a generic
pairs aa or a† a† may be different from zero (anomalous product into a normal form. If the product contains k
correlators). −
factors A+ i mixed with n − k factors Ai , it is:
The variational parameters uk and vk of |BCSi are
chosen to minimize the ground state energy hHi. The N[A± ± P + + −
1 · · · An ] = (±1) Ai1 · · · Aik · · · Ain (14)
evaluation is simplified by Wick’s theorem [2], which is
proven in section III. where for bosons (+1)P = 1, while for fermions (−1)P is
the parity of the permutation that brings the sequence
1 . . . n to the sequence i1 . . . in (another frequently used
C. Independent bosons
notation for normal ordering is : A± ±
1 · · · An :).
For independent bosons the ground state |BECi is a It may appear that normal ordering is not unique, since
Bose-Einstein condensate with N particles in the lowest within + operators or − operators one can choose dif-
energy state, a = 1, and no particles in higher one- ferent orderings. However the different expressions are
+
particle states, at T = 0. Since hBEC|c†1 c√ 1 |BECi = N , actually the same operator, because A operators com-
Bogoliubov
√ ∗ suggested the rescaling c1 = V b and c†1 = mute or anticommute exactly among themselves,+ and
V b , the other operators being left unchanged. Then the same is for A +operators.
−
For example N[A1 A+ 2]
+
can be written as A1 A2 , or with the factors exchanged:
1 N ±A+ +
2 A1 : the two operators coincide.
[b, b∗ ] = , hBEC|b∗ b|BECi = . The action of N -ordering is extended by linearity
V V
from products of components A± i to products of opera-
The operators b and b∗ may be treated as c-numbers tors Ai . For example:
in the thermodynamic limit [4], with |b|2 = N/V . For
− −
any creation and destruction operator, one has the de- N[A1 A2 ] = N[(A+ +
1 + A1 )(A2 + A2 )]
composition into a condensate term, and an excitation = N[A+ + + − − − − +
1 A2 ] + N[A1 A2 ] + N[A1 A2 ] + N[A1 A2 ]
term: + − − − + −
= A+ +
1 A2 + A1 A2 + A1 A2 ± A2 A1
X √
ψi = hi|aica = hi|1i V b + φi , (10)
The following property follows from (14):
a
√
ψi† = hi|ai∗ c†a = hi|1i∗ V b∗ + φ†i
X
(11) N[A1 · · · An ] = (±1)P N[Ai1 · · · Ain ] (15)
a
A product A1 . . . An can be written as a sum of nor-
where φi |BECi = 0 and hBEC|φ†i = 0. mally ordered terms. For two operators the process is

2
straightforward: Lemma IV.2.

A1 A2 =(A+ − + − A−
0 N[A1 · · · An ] (22)
1 + A1 )(A2 + A2 )
n
+ − − − − +
=A+ +
1 A2 + A1 A2 + A1 A2 + A1 A2
X
= N[A−
0 A1 · · · An ] + N[A0 · · · Ai · · · An ].
=N[A1 A2 ] + [A− +
1 , A2 ]∓ (16) i=1

Proof. The proof is by induction. Eq.(22) holds for n =

The last term [A− +
1 , A2 ]∓ is a c-number. The contrac- 1. If, by hypothesis, it holds for n operators, it is now
tion, denoted by a bracket, of two operators is defined proven for n + 1 operators (we write 1± in place of A± 1 ):
as
A1 A2 ≡ A1 A2 − N[A1 A2 ] (17) 0− N[1 · · · (n + 1)]
= 0− 1+ N[2 · · · (n + 1)] + (±1)n 0− N[2 · · · (n + 1)]1−
Combining the contraction definition with (16) yields
= 01 N[ · · · ] ± 1+ 0− N[ · · · ] + (±1)n 0− N[· · · ]1−
A1 A2 = [A− +
1 , A2 ]∓ . (18) The hypothesis of induction is now used in the second
and third terms:
Also, since the gs−expectation value of a normal or-
dered operator is zero, it follows that = N[01 2 · · · (n + 1)] ± 1+ N[0− 2 · · · (n + 1)]
X
A1 A2 = hgs|A1 A2 |gsi = hgs|A1 A2 |gsi (19) ± 1+ N[0· · ·k · · · (n + 1)]
k≥2

The following definition extends the contraction of two + (±)n N[0− 2 · · · (n + 1)]1−
operators to the case where there is a product of n op- X
erators in between: + (±1)n N[0· · ·k · · · (n + 1)]1−
k≥2
0 n 0
A(A1 · · · An )A = (±1) AA (A1 · · · An ) (20) = N[012 · · · (n + 1)] ± N[1+ 0− 2 · · · (n + 1)]
X
IV. WICK’S THEOREM + N[01+ · · ·k · · · (n + 1)]
k≥2

We begin by proving three Lemmas - each one is a + (±1)n N[0− 2 · · · (n + 1)1− ]

generalization of the former. In the first one, a single X
A− operator is at the left of A+ operators, and normal + (±1)n N[02 · · ·k · · · (n + 1)1− ]
ordering is achieved by bringing it to the right of them k≥2

by repeated (anti)commutations. = N[012 · · · (n + 1)] + N[0− 1+ 2 · · · (n + 1)]

Lemma IV.1.
X
+ N[01+ · · ·k · · · (n + 1)]
k≥2
A− + +
0 (A1 · · · An ) = + N[0− 1− 2 · · · (n + 1)]
n
X
N[A− + +
N[A0 · · · Ai · · · A+
X
0 A1 · · · An ] + n] (21) + N[01− 2 · · ·k · · · (n + 1)]
i=1 k≥2
X
Proof. = N[0− 1 · · · (n + 1)] + N[01 · · ·k · · · (n + 1)].
k≥1
A− + +
0 (A1 · · · An )
The last line is (22) with n + 1 operators.
= ([A− + + + + − + +
0 , A1 ]∓ )(A2 · · · An ) ± A1 A0 A2 · · · An
Lemma IV.3.
−
= A0 A1 A+ + + + + +
2 · · · An ± A1 ([A0 , A2 ]∓ )A3 · · · An
+ − + A0 N[A1 · · · An ] (23)
+ A+ +
1 A2 A0 A3 · · · An n
X
= A0 A1 A+ + + + + = N[A0 · · · An ] + N[A0 · · · Ai · · · An ].
2 · · · An + A0 A1 A2 A3 · · · An
i=1
+ − +
+ A+ +
1 A2 A0 A3 · · · An = . . .
n Proof. This is achieved by adding A+
0 N[A1 · · · An ] to
X
A0 A+ + n + + − both sides of Lemma 4.2.
= 1 · · · Ai · · · An + (±1) A1 · · · An A0
i=1
n
Wick’s theorem gives the practical rule to express a
= N[A− +
· · · A+
X
N[A0 · · · Ai · · · A+ product of creation and destruction operators as a sum
0 A1 n] + n ].
i=1
of normally ordered terms. It is an operator identity.
Each contraction, being a c-number, reduces by two the
operator content.

3
Theorem IV.4 (Wick’s Theorem). A1 A2 A3 A4

A1 A2 · · · An = N[A1 · · · An ] (24) = N[1234] + N[1234] + N[1234] + N[1234]

+ N[1234] + N[1234] + N[1234]

X
+ N[A1 · · · Ai · · · Aj · · · An ]
(ij)
+ N[1234] + N[1234] + N[1234]
X
+ N[A1 · · · Ai · · · Ar · · · Aj · · · As · · · An ] = N[1234] + h12iN[34] ± h13iN[24] + h14iN[23]
(ij)(rs) + h23iN[14] ± h24iN[13] ± h34iN[12]
+ ... + h12ih34i ± h13ih24i + h14ih23i (29)

The first sum runs on single contractions of pairs, the

second sum runs on double contractions, and so on. If n An important consequence of Wick’s operator iden-
is even, the last sum contains terms which are products tity is a rule for the expectation value of the product of
of contractions (c-numbers). If n is odd, the last sum an even number of destruction and creation operators:
has terms with single unpaired operators (see examples).
Corollary IV.6.
Proof. The theorem is proven by induction. For n = 2
it is true. Next, suppose that the statement is true for X
a product of creation/destruction operators A1 · · · An : hgs|A1 · · · A2n |gsi = (±1)P hAi1 Aj1 i · · · hAin Ajn i (30)
it is shown that it is true for a product A0 A1 · · · An .
By hypothesis of induction for n operators:
The sum is over all partitions of 1, . . . , 2n into pairs
bn/2c {(i1 , j1 ) . . . (in , jn )} with i# < j# . P is the permutation
X
A1 · · · An = Nn,k (25) that takes 1, . . . , 2n to the sequence i1 , j1 , . . . , in , jn .
k=0

where Nn,k is the sum of normally ordered products of Example IV.7.

n operators with k contractions2 By Lemma 4.3:

hgs|123|gsi = 0 (31)
A0 Nn,k = N [A0 Nn,k ] + N [A0 Nn,k ]. (26)
hgs|1234|gsi = h12ih34i ± h13ih24i + h14ih23i (32)

where A0 Nn,k means the sum of all contractions of A0

with unpaired operators Ai contained in Nn,k . The fol- This is a general rule: two-point correlators deter-
lowing relation takes place: mine all n-point correlators.

N [A0 Nn,k ] + N [A0 Nn,k+1 ] = Nn+1,k+1 . In thermal theory there is no distinguished state to
(27)
define a normal ordering, and thus no Wick’s theorem
Using the induction hypothesis and the last two identi- in the form of an operator identity. Nevertheless, one
ties we find, can prove a thermal analogue of the corollary: for non-
interacting particles: the thermal average of a product
A0 A1 · · · An = A0 Nn,0 + A0 Nn,1 + A0 Nn,2 + . . . of one-particle creation and destruction operators is the
sum of all possible thermal contractions of pairs. The
= N[A0 Nn,0 ] + N[A0 Nn,0 ] + N[A0 Nn,1 ] + N[A0 Nn,1 ] thermal contraction of two operators is the thermal av-
erage of their product[2, 3].
+ N[A0 Nn,2 ] + N[A0 Nn,2 ] + . . .
= Nn+1,0 + Nn+1,1 + . . .

which expresses Wick’s theorem for n + 1 operators. V. WICK’S THEOREM WITH

TIME-ORDERING
Example IV.5.
An important variant of Wick’s theorem deals with
A1 A2 A3 the normal-ordering of a time-ordered product. A nec-
essary condition is that the time evolution of the op-
= N[123] + N[123] + N[123] + N[123]
erators αa and αa+ is a multiplication by some time-
= N[123] + h12iA3 ± h13iA2 + h23iA1 (28) dependent phase factor (c-number). Then, the discus-
sion on normal ordering and contraction of operators
2 For example, Nn,2 =
P
(pq)(rs) N[A1 ...Ap ...Ar ...Aq ...As ...An ]. Ai (ti ) remains unaltered.

4
 A. T-contractions Proof. T(1 · · · n) corresponds to a time ordered se-
quence (±1)P i1 · · · in (we put i = Ai (ti )), to which the
Let us begin with two operators, and apply Wick’s
previous formulation of Wick’s theorem applies :
theorem:
T[12 · · · n] = (±1)P N[i1 · · · in ]
TA1 (t1 )A2 (t2 ) X
+ (±1)P N[i1 · · · i · · · j · · · in ]
= θ(t1 − t2 ){N[A1 (t1 )A2 (t2 )] + A1 (t1 )A2 (t2 )} (ij)

± θ(t2 − t1 ){N[A2 (t2 )A1 (t1 )] + A2 (t2 )A1 (t1 )} X

+ (±1)P N[i1 · · · i · · · r · · · j · · · s · · · in ]
= θ(t1 − t2 )N[A1 (t1 )A2 (t2 )] + θ(t2 − t1 )N[A1 (t1 )A2 (t2 )] (ij)(rs)

+ θ(t1 − t2 )A1 (t1 )A2 (t2 ) ± θ(t2 − t1 )A1 (t1 )A2 (t2 ) + ...
z }| {
= N[A1 (t1 )A2 (t2 )] + A1 (t1 )A2 (t2 ) (33) The factor (±1)P is compensated by restoring the se-
quence 1 · · · n inside the normal ordering. Then the first
The last term is a c-number. We define the time-ordered term is N[1 · · · n]. In doing so, a contraction ij remains
contraction (T-contraction):
unaltered if ti > tj otherwise it is twisted to ji (j pre-
z }| { cedes i in the sequence 1 · · · n). Therefore:
A1 (t1 )A2 (t2 ) = hgs|TA1 (t1 )A2 (t2 )|gsi (34)
z }| {
The T-contraction of two operators with a product of n (±1)P N[i1 · · · i · · · j · · · in ] = N[1 · · · i · · · j · · · n]
operators in between inherits the property of ordinary z}|{
contractions where ij = θ(ti − tj )ij ± θ(tj − ti )ji. This is true for
z }| { z }| { all contractions.
A1 (t1 )(· · · )A2 (t2 ) = (±1)n A1 (t1 )A2 (t2 )(· · · ) (35)
As an interesting application, consider an n-particle
Green function
T-contractions have a new property, not shared by an
ordinary contraction: in G(x1 . . . xn , y1 . . . yn ) (41)
† †
z }| { z }| { = hgs|Tψ(x1 ) . . . ψ(xn )ψ (yn ) . . . ψ (y1 )|gsi
A1 (t1 )A2 (t2 ) = ± A2 (t2 )A1 (t1 ) (36)
where x denotes a complete set of quantum numbers
For field operators we have the explicit expressions: and time (the Heisenberg evolution with the Hamilto-
z }| { nian whose ground state is |gsi). For independent par-
ψ(1)ψ † (2) = hTψ(1)ψ † (2)i = iG0 (1, 2) (37) ticles, Wick’s teorem applies. The average is evaluated
z }| { as a sum of total T-contractions of field operators, i.e.
ψ(1)ψ(2) = hTψ(1)ψ(2)i = iF 0 (1, 2), (38) propagators (we now exclude anomalous propagators):
z }| {
ψ † (1)ψ † (2) = hTψ † (1)ψ † (2)i = iF †0 (1, 2) (39) G0 (x1 . . . xn , y1 . . . yn )
X
= (±1)P G0 (x1 , yi1 ) . . . G0 (xn , yin )
(42)
If |gsi has a definite number of particles, the anomalous
P
correlators F 0 and F †0 are equal to zero. They are non-
zero in the BCS theory. where P is the permutation P (1 . . . n) = (i1 . . . in ).
The sum corresponds to the evaluation of the perma-
B. Wick’s theorem for time-ordered products nent (bosons) or determinant (fermions) of the matrix
G0 (xi , xj ), i, j = 1, . . . n.
For the time-ordered product of several operators,
Wick’s theorem retains the same structure as in (24), Remark V.2. A free theory is a many-particle theory
with T-contractions replacing ordinary ones. The state- where, in some basis, n-particle Green functions are de-
ment is: termined solely by one-particle Green functions.

Theorem V.1 (Wick’ theorem, with time-ordering). Acknowledgement

T[A1 (t1 ) · · · An (tn )] = N[A1 (t1 ) · · · An (tn )] (40) I thank Dr. Daniel Ariad for helpful suggestions that
X z }| { clarified the original text.
+ N[A1 (t1 ) · · · Ai (ti ) · · · Aj (tj ) · · · An (tn )]
(ij)
X REFERENCES
+ N[ · · · double T-contractions · · · ]
[1] G. C. Wick, The evaluation of the Collision Matrix,
+ ... Phys. Rev. 80, 268 (1950)

5
[2] A. L. Fetter and J. D. Walecka, Quantum Theory of
Many Particle Systems (MacGraw Hill, New York,
1971).
[3] M. Gaudin, Une demonstration simplifiee du the-
oreme de Wick en mecanique statistique, Nucl.
Phys. 15, 89-91 (1960).

[4] E. H. Lieb, R. Seiringer and J. Yngvason, Justifica-

tion of c-Number Substitutions in Bosonic Hamil-
tonians, Phys. Rev. Lett. 94, 080401 (2005).