Interaction Representation

So far we have discussed two ways of handling time dependence. The most familiar is
the Schrödinger representation, where operators are constant, and states move in time.
The Heisenberg representation does just the opposite; states are constant in time and
operators move. The answer for any given physical quantity is of course independent of
which is used. Here we discuss a third way of describing the time dependence, introduced
by Dirac, known as the interaction representation, although it could equally well be
called the Dirac representation. In this representation, both operators and states move
in time. The interaction representation is particularly useful in problems involving time
dependent external forces or potentials acting on a system. It also provides a route to
the whole apparatus of quantum field theory and Feynman diagrams. This approach to
field theory was pioneered by Dyson in the the 1950’s. Here we will study the interaction
representation in quantum mechanics, but develop the formalism in enough detail that
the road to field theory can be followed.
We consider an ordinary non-relativistic system where the Hamiltonian is broken up
into two parts. It is almost always true that one of these two parts is going to be handled
in perturbation theory, that is order by order. Writing our Hamiltonian we have

H = H0 + V (1)

Here H0 is the part of the Hamiltonian we have under control. It may not be particularly
simple. For example it could be the full Hamiltonian of an atom or molecule. Neverthe-
less, we assume we know all we need to know about the eigenstates of H0 . The V term
in H is the part we do not have control over. It may be time-dependent, or difficult to
handle in some other way.
In the interaction representation, operators move with H0 . We define
H0 t H0 t
OI (t) = exp(i )OS exp(−i ), (2)
h̄ h̄
where the subscript I means “interaction” and the subscript S means “Schrödinger .”
Now just as Schrödinger and Heisenberg representations must give the same answers, we
must get the same answer from the interaction representation. So we must have

< ΨI (t)|OI (t)|ΨI (t) >=< ΨS (t)|OS |ΨS (t) >, (3)

where for the case where there is no explicit time dependence in H, Schrödinger states
evolve as
Ht
|ΨS (t) >= exp(−i )|ΨS (0) > . (4)
h̄
Since not all the time dependence can be carried by H0 , we must also have an interaction
representation evolution operator which maps interaction states from one time to another.

|ΨI (t2 ) >= UI (t2 , t1 )|ΨI (t1 ) > (5)

Using this and the way Schrödinger states move, we can easily show for the case where
H is constant in time that
i i i
UI (t2 , t1 ) = exp( H0 t2 ) exp(− H(t2 − t1 )) exp(− H0 t1 ). (6)
h̄ h̄ h̄
The operator in the middle is just the evolution operator in the Schrödinger picture.
All this was for the case of no explicit time dependence in H. But there are many inter-
esting problems where the Hamiltonian has explicit time dependence. In the Schrödinger
picture, the operators like X and P are still constant, but things like the potential may
have time dependence. For this case our Hamiltonian could be of the following form

HS = H0 + VS (t). (7)

It should be noted that time dependence does not change any of the basic ideas of
quantum mechanics, or classical mechanics for that matter. The system still exists, and
evolves by unitary transformation from one time to another. The Schrödinger equation
still holds, even though energy will not in general be conserved here. So there is still an
interaction picture evolution operator, but it takes a slightly different form,
i i
UI (t2 , t1 ) = exp( H0 t2 )US (t2 , t1 ) exp(− H0 t1 ), (8)
h̄ h̄
where US is the operator that evolves states in the Schrödinger picture. according to

|ΨS (t2 ) >= US (t2 , t1 )|ΨS (t1 ) > . (9)

The Schrödinger equation still holds,

∂
ih̄ |ΨS (t2 ) >= HS (t2 )|ΨS (t2 ) >, (10)
∂t2
which implies the equation for US ,
∂
ih̄ US (t2 , t1 ) = HS (t2 )US (t2 , t1 ). (11)
∂t2
The other condition US must satisfy is US (t1 , t1 ) = 1. We will not investigate US further,
but proceed directly to studying UI . Using the definition of UI from Eq.(8) and Eq.(11),
we have
∂ i −i
ih̄ UI (t2 , t1 ) = exp( H0 t2 )(−H0 + HS )US (t2 , t1 ) exp( H0 t1 ). (12)
∂t2 h̄ h̄
Using HS = H0 + VS this gives
( )( )
∂ i −i i −i
ih̄ UI (t2 , t1 ) = exp( H0 t2 )(VS (t2 ) exp( H0 t2 ) exp( H0 t2 )US (t2 , t1 ) exp( H0 t1 )
∂t2 h̄ h̄ h̄ h̄
(13)
From the definition of operators in the interaction picture given in Eq.(2) we have finally
∂
ih̄ UI (t2 , t1 ) = VI (t2 )UI (t2 , t1 ), (14)
∂t2
where of course
i −i
VI (t) = exp( H0 t)VS (t) exp( H0 t). (15)
h̄ h̄
To see more explicitly the difference between Schrödinger and interaction picture repre-
sentations of V, suppose that
VS = −XS F (t), (16)
where we are allowing explicit time dependence in F (t), but as always in the Schrödinger
picture XS is constant in time. Now going to the interaction picture, we have
i −i
VI (t) = − exp( H0 t)XS exp( H0 t)F (t) = −XI (t)F (t). (17)
h̄ h̄
In the interaction picture, in addition to the explicit time dependence from F (t), the X
operator also moves with the Hamiltonian H0 .

Perturbation Theory In virtually all cases where the interaction picture is used, a
perturbative expansion in VI is carried out. We expand UI in orders, where “order n”
means that there are n powers of VI in that term. Writing the expansion, we have
∞
∑ (n)
UI (t2 , t1 ) = UI (t2 , t1 ). (18)
n=0

Now one power of VI is a term of order one, so matching orders on left and right sides of
the differential equation for UI given in Eq.(14), we obtain
∂ (n) (n−1)
ih̄ U (t2 , t1 ) = VI (t2 )UI (t2 , t1 ). (19)
∂t2 I
The 0th term is just the identity operator,
(0)
UI (t2 , t1 ) = I. (20)
(0)
Setting UI = I takes care of the boundary condition that UI (t1 , t1 ) = 1, so all subsequent
terms in the expansion must vanish when t1 = t2 . This sets the lower limit of integration
(n)
when we integrate Eq.(19) and we get the integral form for UI given in the following
equation,
i ∫ t2
VI (t′ )UI
(n) (n−1) ′
UI (t2 , t1 ) = − (t , t1 ). (21)
h̄ t1
(n)
To get an explicit formula for UI , it is best to look at a few low order terms. Using
(0)
UI = I, we have ∫
i t2
VI (t′ )dt′
(1)
UI (t2 , t1 ) = − (22)
h̄ t1
Turning to UI2 we have
i ∫ t2
VI (t′2 )UI (t′2 , t1 )dt′2 .
(2) (1)
UI (t2 , t1 ) = − (23)
h̄ t1
(1)
Putting our result for UI in this equation gives

i 2 ∫ t2 ′ ∫ t′2 ′
dt1 VI (t′2 )VI (t′1 ).
(2)
UI (t2 , t1 ) = (− ) dt2 (24)
h̄ t1 t1

At this point, the pattern is clear. Later times are always further to the left. The general
(n)
form for UI is

i ∫ t2 ′ ∫ t′n ′ ∫ t′
dt′1 VI (t′n ) · · · V (t′2 )V (t′1 ).
(n) 2
UI (t2 , t1 ) = (− )n dtn dtn−1 · · · (25)
h̄ t1 t1 t1

(n)
This form of the equation for UI can be used directly in this form. In field theory
applications, it is useful to introduce techniques for writing an equivalent formula in
which all time integrations start and end and the same points, in this case t1 and t2 .

The Forced Harmonic Oscillator A harmonic oscillator acted on by an external

time dependent force is interesting for two reasons. First, it is a model for actual physical
phenomena such as the quantum radiation from a known current. Second, it provides an
excellent case where high order calculations can be carried out analytically in full detail.
The Schrödinger representation full Hamiltonian is
PP 1
HS = + mω 2 XX − XF (t), (26)
2m 2
where
F (t) → 0 |t| → ∞
We will take the term −XF (t), as the “V ” to be used in the interaction representation.
The system will be taken to be in the oscillator ground state at t = −∞, and our goal
will be to obtain the interaction picture state at any subsequent time. This will be given
in terms of the UI operator as follows,

|ΨI (t) >= UI (t, −∞)|0 > . (27)

The interaction picture potential is

VI (t) = −XI (t)F (t), (28)

where XI moves with the oscillator Hamiltonian. In Schrödinger representation, we have

√
h̄
XS = x0 (a + a† ) where x0 = . (29)
2mω
Expressing H0 in terms of creation and destruction operators gives
PP 1 1
H0 = + mω 2 XX = h̄ω(a† a + ). (30)
2m 2 2
Using H0 to determine XI we have
iH0 t −iH0 t
XI (t) = exp( )XS exp( ) = x0 (a exp(−iωt) + a† exp(iωt)) (31)
h̄ h̄
Comparing XI and XS we see that the interaction picture simply supplies motion at
the harmonic oscillator frequency to a and a† . As usual, we can begin to see what is
happening by doing some low order calculations. In first order we have
i∫t
UI1 (t, −∞)|0 > = dt′ x0 F (t′ )(a exp(−iωt′ ) + a† exp(iωt′ ))|0 > (32)
h̄ −∞
i∫t
= dt′ x0 F (t′ ) exp(iωt′ )a† |0 >,
h̄ −∞
where the second equality follows since a|0 >= 0. We see that in first order the system
can only reach the first excited state. Going to second order, we have
i ∫t ∫ t′
dt′2 dt′1 x20 F (t′2 )F (t′1 ).
2
UI2 (t, −∞)|0 >= ( )2 (33)
h̄ −∞ −∞

·(a exp(−iωt′2 ) + a† exp(iωt′2 ))(a exp(−iωt′1 ) + a† exp(iωt′1 ))|0 >

As in the first order case, the destruction operator on the far right destroys |0 > so only
the creation operator on the far right needs to be kept. This puts the system in the first
excited state. The operators on the left can then take the system to the second excited
state, or back to the ground state. Let us focus on the term that ends up in the second
excited state, and use the subscript “excited” for it. We have
∫ t ∫ t′
i ′ ′
dt′2 F (t′2 )eiωt2 dt′1 F (t′1 )eiωt1 a† a† |0 >
(2) 2
(UI (t, −∞)|0 >)excited = ( )2 x20 (34)
h̄ −∞ −∞

This term looks very similar to the first order term “squared” with one obvious difference,
namely the upper limit on the t′1 integration is at t′2 not t. This has a remedy. We note
that the integrand in Eq.(34) is symmetric under the interchange t′1 ↔ t′2 . This means
the integral over the region −∞ < t′1 < t′2 < t is numerically the same as the integral over
the region −∞ < t′2 < t′1 < t. It is useful to draw a simple diagram at this point, showing
the regions of integration with t′2 on the vertical axis and t′1 on the horizontal axis. The
importance of this is that if we now integrate over the region −∞ < t′1 < t, −∞ < t′2 < t,
we have simply doubled the integral, which can be compensated by dividing by 2. We can
then see the beginnings of a coherent state here, which recall would involve the structure
exp(α)t)a† )|0 > . If that is in fact what is happening here, we can see that the coherent
state parameter α(t) must be
i ∫t ′
α(t) = x0 F (t′ )eiωt dt′ . (35)
h̄ −∞
Using this definition, we can write
(0) (1) (2)
UI (t, −∞)|0 > +UI (t, −∞)|0 > +(UI (t, −∞)|0 >)excited (36)
( )
1
= 1 + α(t)a + (α(t)a† )2 |0 >,
†
2
which is the first few terms in the expansion of exp(α)t)a† )|0 > . It is possible using
some more advanced operator techniques to show directly that a coherent state is indeed
the result. Another route is to write a trial form assuming a coherent state, and demon-
strating that this assumed form satifies the correct differential equation. Doing that, we
write
UI (t, −∞)|0 >= exp(Φ(t)) exp(α(t)a† )|0 >≡ |ΨI (t) >, (37)
and demand that |ΨI (t) > satsify

∂
ih̄ |ΨI (t) >= VI (t)|ΨI (t) > .
∂t
where recall that
VI (t) = −x0 F (t)(a exp(−iωt) + a† exp(iωt)).
Given that |ΨI (t) > is a coherent state with parameter α(t), the equation to be verified
is finally
∂
ih̄ |ΨI (t) >= −x0 F (t)(α(t) exp(−iωt) + a† exp(iωt))|ΨI (t) > .
∂t
Determining Φ and showing that this equation is verified is part of Homework Set 9.
The forced oscillator is a case where order by order calculations can be summed up to
give the complete answer. It is worth noting that an oscillator which starts in the ground
state and is acted upon by a quite general force, will remain in a minimum uncertainty
state for all time.