Feynman Path Integral: Formulation of Quantum Dynamics

Masatsugu Sei Suzuki and Itsuko S. Suzuki

Department of Physics, SUNY at Binghamton
(Date: March 03, 2023)

((Overview))
We consider quantum superpositions of possible alternative classical trajectories, a
history being a path in configuration space, taken between fixed points (a and b). The
basic idea of the Feynman path integral is a perspective on the fundamental quantum
mechanical principle of complex linear superposition of such entire spacetime histories.
In the quantum world, instead of there being just one classical ‘reality’, represented by
one such trajectory (one history), there is a great complex superposition of all these
‘alternative realities’ (superposed alternative histories). Accordingly, each history is to be
assigned a complex weighting factor, which we refer to as an amplitude, if the total is
normalized to modulus unity, so the squared modulus of an amplitude gives us a
probability. The magic role of the Lagrangian is that it tells us what amplitude is to be
assigned to each such history. If we know the Lagrangian L, then we can obtain the action
S, for that history (the action being just the integral of L for that classical history along
the path). The complex amplitude to be assigned to that particular history is then given by
the deceptively simple formula amplitude

b
i i
exp( S )  exp(  Ldt ) .
ℏ ℏa

The total amplitude to get from a to b is the sum of these.

((Path integral))
What is the path integral in quantum mechanics?
The path integral formulation of quantum mechanics is a description of quantum
theory which generalizes the action principle of classical mechanics. It replaces the
classical notion of a single, unique trajectory for a system with a sum, or functional
integral, over an infinity of possible trajectories to compute a quantum amplitude. The
basic idea of the path integral formulation can be traced back to Norbert Wiener, who
introduced the Wiener integral for solving problems in diffusion and Brownian motion.
This idea was extended to the use of the Lagrangian in quantum mechanics by P. A. M.
Dirac in his 1933 paper. The complete method was developed in 1948 by Richard
Feynman. This formulation has proven crucial to the subsequent development of
theoretical physics, because it is manifestly symmetric between time and space. Unlike
previous methods, the path-integral allows a physicist to easily change coordinates
between very different canonical descriptions of the same quantum system.

((The idea of the path integral by Richard P. Feynman))

1
 R.P. Feynman, The development of the space-time view of quantum
electrodynamics (Nobel Lecture, December 11, 1965).
http://www.nobelprize.org/nobel_prizes/physics/laureates/1965/feynman-
lecture.html

Feynman explained how to get the idea of the path integral in his talk of the Nobel
Lecture. The detail is as follows. The sentence is a little revised because of typo.
________________________________________________________________
I went to a beer party in the Nassau Tavern in Princeton. There was a gentleman,
newly arrived from Europe (Herbert Jehle) who came and sat next to me. Europeans are
much more serious than we are in America because they think that a good place to
discuss intellectual matters is a beer party. So, he sat by me and asked, «what are you
doing» and so on, and I said, «I’m drinking beer.» Then I realized that he wanted to know
what work I was doing and I told him I was struggling with this problem, and I simply
turned to him and said, ((listen, do you know any way of doing quantum mechanics,
starting with action - where the action integral comes into the quantum mechanics?»
«No», he said, «but Dirac has a paper in which the Lagrangian, at least, comes into
quantum mechanics. I will show it to you tomorrow»
Next day we went to the Princeton Library, they have little rooms on the side to
discuss things, and he showed me this paper. What Dirac said was the following: There is
in quantum mechanics a very important quantity which carries the wave function from
one time to another, besides the differential equation but equivalent to it, a kind of a
kernel, which we might call K ( x ' , x) ,which carries the wave function  (x ) known at
time t, to the wave function  (x' ) at time, t   . Dirac points out that this function K
was analogous to the quantity in classical mechanics that you would calculate if you took
the exponential of i / ℏ , multiplied by the Lagrangian L( xɺ, x) imagining that these two
positions x, x’ corresponded t and t   . In other words, K ( x ' , x) is analogous to
 x ' x
exp[i L( , x)] ,
ℏ 

 x' x
K ( x ' , x )  exp[i L( , x )] . (1)
ℏ 

Professor Jehle showed me this, I read it, he explained it to me, and I said, «what does he
mean, they are analogous; what does that mean, analogous? What is the use of that?» He
said, «you Americans ! You always want to find a use for everything!» I said, that I
thought that Dirac must mean that they were equal. «No», he explained, «he doesn’t
mean they are equal.» «Well», I said, «let’s see what happens if we make them equal.»
So I simply put them equal, taking the simplest example where the Lagrangian is
1
Mxɺ 2  V ( x) , but soon found I had to put a constant of proportionality A in, suitably
2
adjusted. When I substituted exp(iL / ℏ) for K to get

 x' x
 ( x' , t   )   A exp[i L( , x)] ( x, t )dx , (2)
ℏ 

2
and just calculated things out by Taylor series expansion, out came the Schrödinger
equation. So, I turned to Professor Jehle, not really understanding, and said, «well, you
see Professor Dirac meant that they were proportional.» Professor Jehle’s eyes were
bugging out-he had taken out a little notebook and was rapidly copying it down from the
blackboard, and said, «no, no, this is an important discovery. You Americans are always
trying to find out how something can be used. That’s a good way to discover things!» So,
I thought I was finding out what Dirac meant, but, as a matter of fact, had made the
discovery that what Dirac thought was analogous, was, in fact, equal. I had then, at least,
the connection between the Lagrangian and quantum mechanics, but still with wave
functions and infinitesimal times.
It must have been a day or so later when I was lying in bed thinking about these
things, that I imagined what would happen if I wanted to calculate the wave function at a
finite interval later. I would put one of these factors exp(iL / ℏ) in here, and that would
give me the wave functions the next moment, t   and then I could substitute that back
into (2) to get another factor of exp(iL / ℏ) and give me the wave function the next
moment, t + 2, and so on and so on. In that way I found myself thinking of a large
number of integrals, one after the other in sequence. In the integrand was the product of
the exponentials, which, of course, was the exponential of the sum of terms like L / ℏ .
Now, L is the Lagrangian and  is like the time interval dt, so that if you took a sum of
such terms, that’s exactly like an integral. That’s like Riemann’s formula for the integral
 Ldt , you just take the value at each point and add them together. We are to take the
limit as   0 , of course. Therefore, the connection between the wave function of one
instant and the wave function of another instant a finite time later could be obtained by an
infinite number of integrals, (because  goes to zero, of course) of exponential (iS / ℏ)
where S is the action expression (3),

S   L( xɺ , x)dt . (3)

At last, I had succeeded in representing quantum mechanics directly in terms of the

action S. This led later on to the idea of the amplitude for a path; that for each possible
way that the particle can go from one point to another in space-time, there’s an
amplitude. That amplitude is an exponential of i / ℏ times the action for the path.
Amplitudes from various paths superpose by addition. This then is another, a third way,
of describing quantum mechanics, which looks quite different than that of Schrödinger or
Heisenberg, but which is equivalent to them.

1. Introduction
The time evolution of the quantum state in the Schrödinger picture is given by

 (t )  Uˆ (t , t ' )  (t ' ) ,

or

3
 x  (t )  x Uˆ (t , t ' )  (t ' )   dx' x Uˆ (t , t ' ) x' x '  (t ' ) ,

in the x representation, where K(x, t; x’, t’) is referred to the propagator (kernel) and
given by

i
K ( x, t ; x' , t ' )  x Uˆ (t , t ' ) x'  x exp[ Hˆ (t  t ' )] x ' .
ℏ

Note that here we assume that the Hamiltonian Ĥ is independent of time t. Then we get
the form

x  (t )   dx' K ( x, t ; x' , t ' ) x'  (t ' ) .

For the free particle, the propagator is described by

m im( x  x ' ) 2
K ( x, t ; x' , t ' )  exp[ ].
2iℏ(t  t ' ) 2 ℏ( t  t ' )

(which will be derived later)

((Note))
Propagator as a transition amplitude

i
K ( x, t; x ' , t ' )  x exp[  Hˆ (t  t ' )] x '
ℏ
i i
 x exp(  Hˆ t ) exp( Hˆ t ' ) x '
ℏ ℏ
 x, t x ' , t '

Here we define

i i
x, t  exp( Hˆ t ) x , x, t  x exp(  Hˆ t ) .
ℏ ℏ

We note that

i i
x, t a  x exp(  Hˆ t ) a  exp(  Ea t ) x a ,
ℏ ℏ

where

Hˆ a  Ea a .

4
((Heisenberg picture))
The physical meaning of the ket x, t :
The operator in the Heisenberg's picture is given by

i i
xˆ H  exp( Hˆ t ) xˆ exp(  Hˆ t ) ,
ℏ ℏ

i i i
xˆ H x, t  exp( Hˆ t ) xˆ exp( Hˆ t ) exp( Hˆ t ) x
ℏ ℏ ℏ
i ˆ
 exp( Ht ) xˆ x
ℏ
i
 exp( Hˆ t ) x x
ℏ
i
 x exp( Hˆ t ) x  x x, t
ℏ

This means that x, t is the eigenket of the Heisenberg operator x̂H with the eigenvalue
x.
We note that

i
 S (t )  exp(  Hˆ t )  H .
ℏ

Then we get

i
x  S (t )  x exp(  Hˆ t )  H  x, t  H .
ℏ

This implies that

x, t  x, t H
, x  x S.

where S means Schrödinger picture and H means Heisenberg picture.

2. Propagator
We are now ready to evaluate the transition amplitude for a finite time interval

K ( x, t ; x ' , t ' )  x, t x ' , t '

i i i
 x exp[  Hˆ t )] exp[  Hˆ t )]..... exp[  Hˆ t )] x '
ℏ ℏ ℏ

5
where

t  t'
t  (in the limit of N   )
N

t
t0 Dt t

where t0 = t' in this figure.

We next insert complete sets of position states (closure relation)

K ( x, t ; x' , t ' )  x, t x ' , t '

i
  dx1  dx2 .... dx N 2  dx N 1  x exp(  Hˆ t ) x N 1
ℏ
i i
 x N 1 exp(  Hˆ t ) x N 2 ...... x3 exp(  Hˆ t ) x2
ℏ ℏ
i i
 x2 exp(  Hˆ t ) x1 x1 exp(  Hˆ t ) x '
ℏ ℏ

This expression says that the amplitude is the integral of the amplitude of all N-legged
paths.

((Note))

( x ' , t ' ), ( x1 , t1 ), ( x2 , t 2 ), ( x3 , t 3 ), ( x4 , t 4 ),
.....( x N 4 , t N  4 ), ( x N 3 , t N 3 ), ( x N 2 , t N  2 ), ( x N 1 , t N 1 ), ( x, t )

with

t'  t 0  t1  t2  t3  t 4  .......  t N 3  t N 2  t N 1  t  tN

6
We define

x '  x0 , t '  t0 ,
x  xN , t  tN .

We need to calculate the propagator for one sub-interval

i
xi exp[ Hˆ t ) xi 1 ,
ℏ

where i = 1, 2, …, N, and

pˆ 2
Hˆ   V ( xˆ ) ,
2m

Then we have

i i
xi exp( Hˆ t ) xi 1   dpi xi pi pi exp( tHˆ ) xi 1
ℏ ℏ
i
  dpi xi pi pi 1ˆ  tHˆ xi 1  O((t )2 )
ℏ
i pˆ 2
  dpi xi pi ˆ
pi 1  t (  V ( xˆ )) xi 1  O ((t )2 )
ℏ 2m
i pˆ 2 i
  dpi xi pi [ pi 1ˆ  t ( ) xi 1  pi  tV ( xˆ ) xi 1 ]  O ((t )2 )
ℏ 2m ℏ
2
i p i
  dpi xi pi [ pi 1  t ( i ) xi 1  pi  tV ( xi 1 ) xi 1 ]  O((t )2 )
ℏ 2m ℏ
2
i p
  dpi xi pi pi 1  t[ i  V ( xi 1 )] xi 1  O ((t )2 )
ℏ 2m

7
where pi (i = 1, 2, 3, …, N), or

2
i i pˆ
xi exp( Hˆ t ) xi 1   dpi xi pi pi xi 1 [1  t ( i  V ( xi 1 ))]
ℏ ℏ 2m
1 i i

2ℏ  dpi exp[ pi ( xi  xi 1 )[1  tE ( pi , xi 1 )]
ℏ ℏ
1 i i

2ℏ  dpi exp[ pi ( xi  xi 1 )] exp[ tE ( pi , xi 1 )]
ℏ ℏ
1 i ( x  xi 1 )

2ℏ  dpi exp[ { pi i
ℏ t
t  E ( pi , xi 1 )t}]

1 i ( x  xi 1 )

2ℏ  dpi exp[ { pi i
ℏ t
 E ( pi , xi 1 )}t ]

where

pi 2
E( pi , x i1 )   V(xi 1 ) .
2m

Then we have

K ( x, t ; x ' , t ' )  x, t x ' , t '

dp1 dp2 dp dp
 lim  dx1  dx2 ....... dx N 1  ....... N 1  N
N  2ℏ 2ℏ 2ℏ 2ℏ
( x  xi 1 )
N 2
i p
 exp[ { pi i  ( i  V ( xi 1 ))}t ]
ℏ i 1 t 2m

We note that

dpi i ( xi  xi 1 ) pi 2 dpi i pi 2
 2 ℏ ℏ i t
exp[ { p 
2m
}t ]   2 ℏ ℏ i i i1 2mℏ t ]
exp[ { p ( x  x )  i

m imt xi  xi 1 2
 exp[ ( ) ]
2 ℏit 2ℏ t

_____________________________________________________________________
((Mathematica))

8
 p2
Clear@"Global`∗"D; f1 = ExpB p x − ∆t F;
1
2π — — 2m—
Integrate@f1, 8 p, −∞, ∞<D êê
Simplify@ , 8— > 0, m > 0, Im @∆tD  0<D &
m x2
2 ∆t —

∆t —
2π
m
______________________________________________________________________
Then we have

m N /2
K ( x, t ; x ' , t ' )  lim  dx1  dx2 ....... dx N 1 ( )
N  2ℏit
i N
m x  xi 1 2
 exp[ t { ( i )  V ( xi 1 )}]
ℏ i 1 2 t

Notice that as N   and therefore t  0 , the argument of the exponent becomes the
standard definition of a Riemann integral

m x  xi 1 2
N t
i i
lim t  { ( i )  V ( xi 1 )}   dtL( x, xɺ ) ,
N  ℏ
t  0 i 1 2 t ℏ t'

where L is the Lagrangian (which is described by the difference between the kinetic
energy and the potential energy). Mathematically, we had better to use
t t
i i

ℏ t'
dtL( x, xɺ )   dt1 L( x(t1 ), xɺ (t1 )
ℏ t'

9
 x

O t

1
L( x, xɺ )  m( xɺ ) 2  V ( x) .
2

It is convenient to express the remaining infinite number of position integrals using the
shorthand notation

m
 D[ x(t )]  lim  dx  dx ....... dx
N 
1 2 N 1 (
2ℏit
)N / 2 .

Thus we have

i
K ( x, t ; x ' , t ' )  x, t x' , t '   D[ x (t )] exp{ S [ x(t )]} ,
ℏ

where

t
S [ x (t )]   dtL( x , xɺ ) .
t'

The unit of S is [erg sec].

10
 When two points at (ti, xi) and (tf, xf) are fixed as shown the figure below, for
convenience, we use

tf

S [ x (t )]   dtL( x, xɺ ) .
ti

x2

x1

t
t1 t2

This expression is known as Feynman’s path integral (configuration space path integral).
S [ x (t )] is the value of the action evaluated for a particular path taken by the particle. If
one wants to know the quantum mechanical amplitude for a point particle at x’, at time t’
to reach a position x, at time t, one integrates over all possible paths connecting the points
with a weight factor given by the classical action for each path. This formulation is
completely equivalent to the usual formulation of quantum mechanics.
The expression for K ( x, t; x ' , t ' )  x, t x ' , t ' may be written, in some loose sense, as

iS ( N ,0)
x N  x, t N  t x0  x' , t 0  t '  
( all ) path
exp[
ℏ
]

iS path 1 iS path  2 iS path  n

 exp( )  exp( )      exp( )  
ℏ ℏ ℏ

11
where the sum is to be taken over an innumerably infinite sets of paths.

xcl Ht L

(a) Classical case

Suppose that ℏ  0 (classical case), the weight factor exp[iS / ℏ] oscillates very
violently. So there is a tendency for cancellation among various contribution from
neighboring paths. The classical path (in the limit of ℏ  0 ) is the path of least action,
for which the action is an extremum. The constructive interference occurs in a very
narrow strip containing the classical path. This is nothing but the derivation of Euler-
Lagrange equation from the classical action. Thus the classical trajectory dominates the
path integral in the small ħ limit.
In the classical approximation ( S  ℏ )

iS cl
x N  x, t N  t x0  x' , t 0  t '  "smooth function" exp( ). (1)
ℏ

But at an atomic level, S may be compared with ħ, and then all trajectory must be added
in x N  x, t N  t x0  x ' , t 0  t ' in detail. No particular trajectory is of overwhelming
importance, and of course Eq.(1) is not necessarily a good approximation.

12
(b) Quantum case.
What about the case for the finite value of S / ℏ (corresponding to the quantum case)?
The phase exp[iS / ℏ] does not vary very much as we deviate slightly from the classical
path. As a result, as long as we stay near the classical path, constructive interference
between neighboring paths is possible. The path integral is an infinite-slit experiment.
Because one cannot specify which path the particle choose, even when one knows what
the initial and final positions are. The trajectory can deviate from the classical trajectory
if the difference in the action is roughly within ħ.

((REFERENCE))
R.P. Feynman and A.R. Hibbs, Quantum Mechanics and Path Integrals (extended edition)
Dover 2005.

((Note))
We can use the Baker-Campbell-Hausdorff theorem for the derivation of the
Feynman path integral. We have a Hamiltonian

Hˆ  Tˆ  Vˆ

where Tˆ is the kinetic energy and Vˆ ( xˆ ) is the potential energy. We consider

i i
exp( Hˆ t )  exp[ (Tˆ  Vˆ )t ]
ℏ ℏ
 exp( P  Qˆ )
ˆ

where

i i
Pˆ   Tˆ t , Qˆ   Vˆ t
ℏ ℏ

We use the Baker-Campbell-Hausdorff theorem

1 1 1
exp( Pˆ ) exp(Qˆ )  exp{Pˆ  Qˆ  [ Pˆ , Qˆ ]  [ Pˆ ,[ Pˆ , Qˆ ]]  [Qˆ ,[ Pˆ , Qˆ ]]  ...}
2 12 12

with

i
[ Pˆ , Qˆ ]  ( )2 (t )2 [Tˆ , Vˆ ]
ℏ

i
[ Pˆ ,[ Pˆ , Qˆ ]]  ( )3 (t )3[Tˆ ,[Tˆ , Vˆ ]]
ℏ

13
 i
[Qˆ ,[ Pˆ , Qˆ ]]  ( )3 (t )3[Vˆ ,[Tˆ , Vˆ ]]
ℏ

In the limit of t  0 , we get

i
exp( Hˆ t )  exp( Pˆ  Qˆ )
ℏ
 exp( Pˆ ) exp(Qˆ )
i i
 exp( Tˆ t ) exp( Vˆ t )
ℏ ℏ

Using this, we get the matrix element

i i i
x j exp( Hˆ t ) x j 1  x j exp( Tˆ t ) exp( Vˆ t ) x j 1
ℏ ℏ ℏ
i t 2 i
 x j exp( pˆ ) x j 1 exp( V ( x j 1 )t )
2mℏ ℏ

or

i m ( x j 1  x j )
1/2 2
i ˆ  m 
x j exp( H t ) x j 1    exp{ [  V ( x j 1 )]t )]
ℏ  2 iℏt  ℏ 2 (t )2

3. Free particle propagator

In this case, there is no potential energy.

m N/2 i N
m x x
K ( x, t; x' , t ' )  lim  dx1  dx2 ....... dx N 1 ( ) exp[ t { ( i i 1 ) 2}] ,
N  2ℏit ℏ i 1 2 t

or

K(x,t;x' ,t' )  lim  dx1  dx2 ....... dx N 1

N 

m m
( ) N / 2 exp[ {(x1  x' )2  (x 2  x1 )2  (x 3  x 2 )2  ....  (x  x N 1 )2}]
2ℏit 2 ℏit

We need to calculate the integrals,

m 2/ 2  m
f1  ( )  dx1 exp[ {( x1  x' ) 2  ( x 2  x1 ) 2 ]
2ℏit 
2ℏ i  t
1 m 1/ 2 m
 ( ) exp[ ( x 2  x' ) 2 ]
4 ℏ i t 4ℏ i  t

14
 m 1/ 2 m
g1  f 1 ( ) exp[ ( x3  x2 ) 2 ]
2ℏit 2ℏit

1 m 1/ 2 m
f2   g dx
1 2  (
6 ℏit
) exp[ (x  x' ) 2 ]
6ℏit 3


m 1 /2 m
g2  f2 ( ) exp[ (x4  x3 )2 ]
2ℏit 2ℏit

1 m 1/ 2 m
f3   g2 dx 3  ( ) exp[ (x 4  x' )2 ]

8 ℏit 8ℏit

..........................................................................................................................................

m m(x  x' )2
K(x,t;x' ,t' )  lim ( 1/ 2
) exp[ ],
N  2 ℏNit 2ℏiNt

or

m  m(x  x' )2
K(x,t;x' ,t' )  (
1/ 2
) exp[ ],
2ℏi(t  t' ) 2ℏi(t  t' )

where we uset  t'  Nt in the last part.

((Mathematica))

15
Free particle propagator;
e=Dt

Clear@"Global`∗"D;
exp_ ∗ := exp ê. 8Complex@re_, im_D Complex@re , − im D<;
h1 = IHx1 − x 'L + Hx2 − x1L M êê Expand;
2 2

2
2π — ε −2 −m
f0@x1_D := ExpB h1F;
m 2 —ε

f1 = Integrate@f0@x1D, 8x1, −∞, ∞<D êê

SimplifyB , Im B F > 0F &
m
ε—
m Hx2−x′ L2
4ε— − m
ε—

2 π

1
2π — ε −2 −m
g1 = f1 ExpB Hx3 − x2L2 F êê Simplify;
m 2 —ε

f2 = ‡ g1 x2 êê SimplifyB , Im B F > 0F &

∞ m
−∞ ε—
m Hx3−x′ L2
6ε—

ε—
6π
m

1
2π — ε −2 −m
g2 = f2 ExpB Hx4 − x3L2 F êê Simplify;
m 2 —ε

f3 = ‡ g2 x3 êê SimplifyB , Im B F > 0F &

∞ m
−∞ ε—
m Hx4−x′ L2
8ε— − m
ε—

2 2π

1
2π — ε −2 −m
g3 = f3 ExpB Hx5 − x4L2 F êê Simplify;
m 2 —ε

f4 = ‡ g3 x4 êê SimplifyB , Im B F > 0F &

∞ m
−∞ ε—
m Hx5−x′ L2
10 ε —

ε—
10 π
m
16
4. Gaussian path integral
The simplest path integral corresponds to the vase where the dynamical variables
appear at the most up to quadratic order in the Lagrangian (the free particle, simple
harmonics are examples of such systems). Then the probability amplitude associated with
the transition from the points ( xi , t i ) to ( x f , t f ) is the sum over all paths with the action
as a phase angle, namely,

i
K ( x f , t f ; xi , t i )  exp[ S cl ]F (t f , ti ) ,
ℏ

where S cl is the classical action associated with each path,

tf

S cl   dtL( xcl , xɺcl , t ) ,

ti

with the Lagrangian L( x, xɺ, t ) described by the Gaussian form,

L( x, xɺ , t )  a(t ) xɺ 2  b(t ) xɺx  c(t ) x 2  d (t ) xɺ  e(t ) x  f (t )

If the Lagrangian has no explicit time dependence, then we get

F (t f , ti )  F (t f  t i ) .

For simplicity, we use this theorem without proof.

i
K ( x f , t f ; xi , ti )  exp[ Scl ]F (t f  ti ) .
ℏ

((Proof)) R.P. Feynman and A.R. Hibbs, Quantum Mechanics and Path Integrals.

Let xcl (t ) be the classical path between the specified end points. This is the path which is
an extremum for the action S. We can represent x(t ) in terms of xcl (t ) and a new
function y (t ) ;

x(t )  xcl (t )  y (t ) ,

where y (ti )  y (t f )  0 . At each time t, the variables x(t ) and y (t ) differ by the
constant xcl (t ) (Of course, this is a different constant for each value of t). Thus, clearly,

dxi  dyi ,

17
Fig. x(ti )  xcl (ti )  y (ti ) . xcl (ti ) is constant for ti  t  ti  t  ti 1 , which is
independent of the path, while x(ti ) and y (ti ) are also constant for
ti  t  ti  t  ti 1 , but depends on the path chosen. So that we have
d x(ti )  dy (ti ) which depends on the choice of path.

for each specific point ti in the subdivision of time. In general, we may say that

Dx (t )  Dy (t ) .

The integral for the action can be written as

tf

S [t f , ti ]   L[ xɺ (t ), x (t )), t ]dt ,
ti

with

L( xɺ, x, t )  a (t ) xɺ 2  b(t ) xɺx  c(t ) x 2  d (t ) xɺ  e(t ) x  f (t ) .

We expand L( xɺ, x, t ) in a Taylor expansion around xcl , xɺcl . This series terminates after
the second term because of the Gaussian form of Lagrangian. Then

18
 L L 1 2 L 2L 2L
L( xɺ, x, t )  L( xɺcl , xcl , t )  | xcl y  | xɺcl yɺ  ( 2 y 2  yyɺ  2 yɺ 2 ) | xcl , xɺcl .
x xɺ 2 x xxɺ xɺ

From here we obtain the action

tf
L L
S [t f , ti ]  Scl [t f , ti ]   dt ( | xcl y  | x yɺ )
ti
x xɺ cl
tf
1 2 L 2 2 L 2L 2
2 ti
 dt ( y  ɺ
y y  y ) | xcl , xɺcl
xɺ 2 xɺx x 2
tf
L L
 Scl [t f , ti ]   dt ( | xcl y  | x yɺ )
ti
x xɺ cl
tf

  dt[a (t ) yɺ 2  b(t ) yyɺ  c (t ) y 2 ]

ti

The integration by parts and use of the Lagrange equation makes the second term on the
right-hand side vanish. So, we are left with

tf

S [t f , ti ]  S cl [t f , ti ]   dt[a (t ) yɺ 2  b(t ) yyɺ  c(t ) y 2 ] .

ti

Then we can write

tf

S [ x (t )]  Scl [t f , ti ]   [a (t ) yɺ 2  b(t ) yɺy  c(t ) y 2 ]dt .

ti

The integral over paths does not depend on the classical path, so the kernel can be written
as

y 0 tf
i i
K ( x f , t f ; xi , ti )  exp[ Scl ]  exp{  [a(t ) yɺ 2  b(t ) yɺy  c(t ) y 2 ]dt}Dy (t )
ℏ y 0
ℏ ti
i
 F (t f , ti ) exp[ Scl ]
ℏ

where

y 0 tf
i
F (t f , ti )   exp{  [a (t ) yɺ 2  b(t ) yɺy  c (t ) y 2 ]dt}Dy (t ) .
y 0
ℏ ti

19
It is defined that F (t f , ti ) is the integral over all paths from y  0 back to y  0 during
the interval (t f  ti ) .

REFERENCES
S.Rajasekar and R. Velusamy, Quantum Mechanics II, Advanced Topics (CRC Press,
2015).

((Note))
If the Lagrangian is given by the simple form

L( x, xɺ , t )  a(t ) xɺ 2  b(t ) xɺx  c (t ) x 2

then F (t f , ti ) can be expressed by

F (t f , ti )  K ( x f  0, t f ; xi  0, ti ) .

4. Evaluation of F (t f , ti ) for the free particle

We now calculate F (t f  t , ti  t ' ) for the free particles, where the Lagrangian is
given by the form,

1 2
L( xɺ, x, t )  mxɺ .
2

Then we have

y 0 tf
i m
F (t f , ti )   exp{  yɺ 2dt}Dy (t )
y 0
ℏ ti 2 .

Replacing the variable y by x, we get

x 0 tf
i m
F (t f , ti )   exp{  xɺ 2dt}Dx(t )
x 0
ℏ ti 2

In this case, formally F (t f , ti ) is equal to the propagator K ( x f  0, t f ; xi  0, ti ) ,

20
 F (t f , ti )  K ( x f  0, t f ; xi  0, ti )
 lim  dx1  dx2 ....... dxN 1
N 

m N /2 m
{x1  ( x2  x1 ) 2  ( x3  x2 ) 2  ....  ( x N 1  xN  2 )2  xN 1 }]
2 2
( ) exp[
2ℏit 2ℏit

t f  ti t  t'
where we put xi  x ' 0 , and x f  x  0 , and t   .
N N
We now evaluate the following integral;

F (t f , ti )  lim  dx1  dx2 ....... dxN 1

N 

m m
{x1  ( x2  x1 ) 2  ( x3  x2 ) 2  ....  ( x N 1  x N 2 ) 2  x N 1 }]
2 2
( ) N / 2 exp[
2ℏit 2ℏit

We note that

f  x1  ( x2  x1 ) 2  ( x3  x2 )2  ....  ( xN 1  xN  2 )2  xN 1
2 2

 2( x1  x2  ....  x N 1 )  2( x1 x2  x2 x3  ....  xN  2 x N 1 )
2 2 2

Using the matrix, f can be rewritten as

f  X  Aˆ X  (Uˆ )  AUˆ   Uˆ  Aˆ Uˆ   T (Uˆ  Aˆ Uˆ ) ,

with

 x1   2 1 0 0 0 . 0
   
 x2  1 2 1 0 0 . . 
 .   0 1 2 1 0 . . 
   
X  . , A   0 0 1 2 1 . . .
   .
 .   . . . . . . 
 .   . . . . . 2  1
   
 xN 1   . . . . . 1 2 

We solve the eigenvalue problems to determine the eigenvalues and the unitary operator,
such that

21
  1 0 0 0 . . 0
 
 0 2 0 0 . . 0
0 0 3 0 . . 0
 
Uˆ  Aˆ Uˆ   0 0 0 4 . . 0 ,
 
. . . . . . . 
. . . . . . . 
 
0 0 0 0 . . n 

where i is the eigenvalue of A. Then we have

 1 0 0 0 . . 0  1 
  
 0 2 0 0 . . 0   2 
0 0 3 0 . . 0  . 
  
f  1 1 1 1 . .  N 1  0 0 0 4 . . 0  . 
. . . . . . .  . 

. . . . . . .  . 
  
0 0 0 0 . . n  N 1 
N 1
  i i
2

i 1

The Jacobian determinant is obtained as

 ( x1, x2 ,..., x N 1 )
 det U  1 .
 (1 , 2 ,..., N 1 )

Then we have the integral

22
 F (t f , ti )  F (t , t ' )
m N /2 m
 lim  d1  d 2 ....... d N 1 ( {11  2 2  ...  N 1 N 1 }]
2 2 2
) exp[
N  2ℏit 2ℏit
m N / 2 2ℏit 2ℏit 2ℏit
( ) ...
2ℏit m1 m2 mN 1
m N/2 m  ( N 1) / 2 1
( ) ( )
2ℏit 2ℏit 12 ....N 1.
m

2ℏit (det A)
m

2ℏiNt
m

2ℏi (t  t ' )

where

det A  123... N 1  N

((Mathematica)) Example (N = 6).

The matrix A (5 x 5): N  1  5

2 1 0 0 0
 
1 2 1 0 0
A   0 1 2 1 0
 
 0 0 1 2 1
 0 0 0 1 2 


The eigenvalues of A:

1  2  3 , 2  3 , 3  2 , 4  1 , 5  2  3 .

The unitary operator

Û 

23
 N 1
f   ii .
2

i 1

 1 0 0 0 0
 
 0 2 0 0 0
Uˆ  Aˆ Uˆ   0 0 3 0 0 .
 
0 0 0 4 0

0 0 0 0 5 

det A  N  6 .
________________________________________________________________________
__

24
Clear "Global` " ;

2 1 0 0 0
1 2 1 0 0
A1 0 1 2 1 0 ;
0 0 1 2 1
0 0 0 1 2
eq1 Eigensystem A1

2 3 , 3, 2, 1, 2 3 , 1, 3 , 2, 3, 1 ,
1, 1, 0, 1, 1 , 1, 0, 1, 0, 1 ,
1, 1, 0, 1, 1 , 1, 3 , 2, 3, 1

1 Normalize eq1 2, 1 Simplify

1 1 1 1 1
, , , ,
2 3 2 3 2 2 3

2 Normalize eq1 2, 2 Simplify

1 1 1 1
, , 0, ,
2 2 2 2

25
 3 Normalize eq1 2, 3 Simplify
1 1 1
, 0, , 0,
3 3 3

4 Normalize eq1 2, 4 Simplify

1 1 1 1
, , 0, ,
2 2 2 2

5 Normalize eq1 2, 5 Simplify

1 1 1 1 1
, , , ,
2 3 2 3 2 2 3

UT 1, 2, 3, 4, 5 ;U Transpose UT ;
UH UT;
U MatrixForm
1 1 1 1 1
2 3 2 3 2 2 3
1 1 1 1
0
2 2 2 2
1 1 1
0 0
3 3 3
1 1 1 1
0
2 2 2 2
1 1 1 1 1
2 3 2 3 2 2 3

26
 UH.U Simplify; UH.U MatrixForm
1 0 0 0 0
0 1 0 0 0
0 0 1 0 0
0 0 0 1 0
0 0 0 0 1

1
2
3 ;
4
5

s1 UH.A1.U Simplify; s1 MatrixForm

2 3 0 0 0 0
0 3 0 0 0
0 0 2 0 0
0 0 0 1 0
0 0 0 0 2 3

f1 Transpose .s1. FullSimplify

2 3 12 3 22 2 32 42 2 3 52

Det A1
6

K1 eq1 1, 1 eq1 1, 2 eq1 1, 3 eq1 1, 4

eq1 1, 5 Simplify
6

6. Calculation of the average  Â 

In order to understand the above discussion, for the sake of clarity, we discuss the
fundamental mathematics in detail.
6.1 The average  Â  under the original basis { bi }

27
 We consider the two bases { bi , ai }, where the new basis { ai } is related to the

original basis {} bi through a unitary operator Û ,

a j  Uˆ b j , b j  Uˆ  a j , b j  a j Uˆ ,

with Uˆ Uˆ  1ˆ . ai is the eigenket of the Hermitian operator  with the eigenvalue

ai .

Aˆ ai  ai ai .

Note that

bi a j  bi Uˆ b j  ai Uˆ a j ,

or

ai Uˆ a j  bi Uˆ UU
ˆ ˆ b  b Uˆ b .
j i j

In other word, the matrix element of Û is independent of the kind of basis (this is very
important property). We also note that

ai Aˆ a j  bi Uˆ  AU
ˆ ˆ b  a
j i ij (diagonal matrix)

Here we define the Column matrices for the state  of the system,

28
  1   1 
   
 2   2 
 .   . 
β  α   (matrix form)
 .   . 
 .   . 
   
 n  n 

with

i  bi  ,  i  ai  .

We now consider the average over the state  under the original basis { bi }.

A   Aˆ 
   bi bi Aˆ b j bi 
i, j

  bi 
*
bi Aˆ b j bi 
i, j

   i * Aij i
i, j

 A11 A12 . . . A1n  1 

  
 A21 A22 . . . An   2 
 . . . . . .  . 
  1*  2* . . .  n*    
 . . . . . .  . 
 . . . . . .  . 
 
  
 An1 An 2 . . . Ann  n

 β  Aβ

(matrix form), using the closure relation. The relation between β and α is obtained as
follow.

29
  i  ai 
  ai b j b j 
j

  ai b j  j
j

  bi Uˆ  b j  j
j

or

α  Uβ (matrix form)

since ai  bi Uˆ  .

6.2 The average  Â  under the new basis { ai }

Next, we now consider the average under the new basis { ai }.

A   Aˆ 
   ai ai Aˆ a j a j 
i, j

  ai 
*
ai Aˆ a j aj 
i, j

   i* ai ij i
i, j

 a1 0 . . . 0  1 
  
0 a2 . . . 0   2 
 . . . . . .  . 
 1*  2* . . .  n*    
 . . . . . .  . 
 . . . . . .  . 
 
  
0 0 . . . an  n

  ai  i
2

i, j

6.3 The calculation of the average using Mathematica

(i) Find eigenvalue and eigenkets of matrix A by using Mathematica

30
 Eigensystem[A]

which leads to the eigenvalues ai and eigenkets, ai . The eigenkets ahould be

normalized using the program Normalize. When the system is degenerate ( the same
eigenvalues but different states), further we need to use the program Orthogonize for all
eigenkets obtained by doing the process of Eigensystem[A]

(ii) Determine the unitary matrix U

Unitary matrix U is defined as

 b1 a1 b1 a2 . . . b1 an 
 
 b2 a1 b2 a2 . . . b2 an 
 . . . . . . 
U 
 . . . . . . 
 . . . . . . 
 
 b a bn a2 . . . bn an 
 n 1
  u1 u 2 . . . u n 

where

 b1 ai 
 
 b2 ai 
 b3 ai 
ui    (matrix form of eigenkets)
 . 
 . 
 
 bn ai 


Thus, we have

α  Uβ , β  Uα (matrix form)

31
where

U   u1 u 2 ......u n 

and

 u1* 
 *
 u2 
 . 
U   
 . 
 . 
 *
u 
 n 

Note that

β  Aβ  (α  U
ˆ  AUα )  (α  Aα
ɶ )

where

 a1 0 . . . 0
 
 0 a2 . . . 0
 . 
ˆ  AU   .
U
. . . .
 (diagonal matrix)
. . . . . . 
. . . . . . 
 
0 0 . . . an 

6.4 Example-1 (3x3 matrix)

Here we discuss a typical example, A is 3x3 matrix.

32
 f  2(12   2 2  3 2  1  2   2  3   3 1 )
 β  Aβ
 1 
  1*  2* 3*  A  2 
 
 3
 1 
  1 2 3  A  2 
 
 3

Where 1 ,  2 , and  3 are real,

 1 
 
β   2  , β    1*  2* 3*    1 2 3   βT ,
 
 3

 2 1 1
 
A   1 2 1  .
 1 1 2 
 

Eigenvalue problem of matrix A (we solve the problem using Mathematica. The system
is degenerate)

A1  a11 , A2  a22 , A3  a33 .

where the eigenvalues and eigenkets are as follows,

1
1  
a1  3 , 1  0 ,
2  
 1

33
 1
1  
a2  3 , 2  2 ,
6  
1

 1
1  
a3  0 , 3  1 .
3  
 1

under the basis { bi } . The unitary matrix can be obtained as

 1 
 
U  1 2 3  U    2  
  
 1 1 1   3 
 
 2 6 3  1 1 
 0 
 2 1 , 2 2
 0   
 6 3  1 2 1 
  
 1 1 1  6 6 6
  
 2 6 3  1 1 1 
 
 3 3 3

 3 0 0
   
U U  1, U AU   0 3 0 
 0 0 0
 

 1 1 1   1 1 1 
   1  2  3 
 1   2 6 3 
 1 
2 6 3 
   2 1    2 1 
β    2   Uα   0   2     2  3 
   6 3    6 3 
 3 
 1 1 1  3   1 1 1 
   1   2  3 
 2 6 3  2 6 3 

Thus, we have

34
 f  β  Aβ
 α  (U  AU)α
 a1 0 0   1 
  1
*
2 *
   0
3
*
a2 0    2 
0 0 a3   3 

2 2 2
 a1 1  a2  2  a3 3
 312  3 2 2  0 32

6.5 Example-2 4x4 matrix

We also discuu the second example; A is 4x4 matrix.

f  12  2  2 2  2  3 2   4 2  2 1  2  2  2  3  2  3 1
 β  Aβ
 1 
 
  1*  2* 3  4  A  2 
* *
 3 
 
 4 
 1 
 
  1 2 3  4  A  2 
 3 
 
 4 

Where 1 ,  2 ,  3 and  4 are real,

 1 
 
2
β  β    1*  2* 3*  4*    1  2 3  4   βT
 3 
 
 4 

35
  1 1 0 0 
 1 2 1 0 
A  .
 0 1 2 1
 
 0 0 1 1 

Eigenvalue problem of matrix A (using Mathematica)

A1  a11 , A2  a22 ,

A3  a33 , A4  a44

The eigenvalues and eigenkets are obtained as follows,

 1 
 
 2 2 2 
 1 1 
 1 
 2 2 
a1  2  2 , 1   ,
 1 1
1 
 2 2 
 
 1 
 
 2 2 2 

 1 
 2 
 
 1 
 
a2  2 , 2   2  ,
 1 
 2
 1 
 
 2 

36
  1 1 
 1 
 2 2 
 1 1 
 1 
 2 2 
a3  2  2 , 3   .
1 1
1 
 2 2
 
 1 1 
 2 1 
 2

1
2
 
1
 
a4  0 , 4   2  .
1
2
1
 
2

The unitary matrix:

U  1 2 3 4 
 1 1 1 1 1
 1 
 2 2 2 2 2 2 2
 
 1 1
1

1 1
1
1 1
 2 2 2 2 2 2 ,
 
 1 1 1 1 1 1
 2 1   1
2 2 2 2 2
 
 1 1 1 1 1
  1 
 2 2 2 2 2 2 2 


37
  1 
 

U   2 

 3 
  
 4 
 1 1 1 1 1 1 
  1 1  
 2 2 2 2 2 2 2 2 2 2 
 1 1 1 1 
   
 2 2 2 2 
1 1 1 1 1 1 1 1 
 1 1  1  1 
2 2 2 2 2 2 2 2
 
 1 1 1 1 
 2 2 2 2 

2 2 0 0 0
 
0 2 0 0
U U  1 , U AU  

,
 0 0 2  2 0
 0 0 0 0 


 1 1 1 1 1
 1 
 2 2 2 2 2 2 2
 
 1   1 1 1 1 1 1   1 
  1  1  
 2 2 2 2 2 2  2 
β   2   Uα    ,
 3   1 1 1 1 1 1   3 
   2 1   1  

 4 2 2 2 2 2  4 
 
 1 1 1 1 1
   1 
 2 2 2 2 2 2 2 

or

38
  1 1 1 1 1 
 1   2  1 3   4 
 2 2 2 2 2 2 2 
 
 1   1 1 1 1 1 1 
     1 1   2  1 3   4
 2 2 2 2 2 2 2 
.
 3   1 1 1 1 1 1 
   1 1   2  1 3   4 
 4   2 2 2 2 2 2

 1 1 1 1 1 
  1   2  1 3   4 
 2 2 2 2 2 2 2 

Thus, we have

f  β  Aβ
 α  (U  AU)α
2 2 0 0 0   1 
  
*  0 2 0 0  2 
 1*  2*  3* 4   
 0 0 2  2 0   3 
 0  
 0 0 0    4 
2 2 2 2
 a1 1  a2  2  a3  3  a4  4
 (2  2)12  2 2 2  (2  2) 3 2  0 4 2

7. Equivalence with Schrödinger equation

The Schrödinger equation is given by


iℏ  (t )  Hˆ  (t ) .
t

For an infinitesimal time interval  , we can write

i ˆ
 ( )   (0)   H  ( 0) ,
ℏ

from the definition of the derivative, or

39
 i
x  ( )  x  (0)   x Hˆ  (0)
ℏ
i ℏ2 2
  [  V ( x)] x  (0)
ℏ 2m x 2

or

i
 ( x,  )  ( x,0)   x Hˆ  (0)
ℏ
i ℏ2 2
  [  V ( x)] ( x,0)
ℏ 2m x 2

in the x representation.
We now show that the path integral also predicts this behavior for the wave function.
To this end, we start with


x  ( )   dx' K ( x,  ; x' ,0) x'  (0) ,


or


 ( x,  )   dx ' K ( x,  ; x' ,0) ( x' ,0) ,


where

m i x  x ' x  x '
K ( x,  ; x' ,0)  exp[ L( , )]
2iℏ ℏ  2
m i 1 ( x  x ' ) 2 x  x'
 exp[ { m V ( )}]
2iℏ ℏ 2  2
2

Then we get


m i 1 ( x  x ' ) 2 x  x'
 ( x,  ) 
2iℏ  dx' exp[

{ m
ℏ 2  2
V (
2
)}] ( x' ,0) .

We now define

x' x   .

Then we have

40
 
m im 2 i 
 ( x,  ) 
2iℏ  d exp[

2ℏ 
 V ( x  )] ( x   ,0) .
ℏ 2

The dominant contribution comes from the small limit of . Using the Taylor expansion
in the limit of   0


V ( x  )  V ( x ) ,
2

 ( x,0)  2  2 ( x,0)
 ( x   ,0)   ( x,0)    ,
x 2 x 2

we get


m im 2 i  ( x,0)  2  2 ( x,0)
 ( x,  ) 
2iℏ  d exp(
 2ℏ 
)[1  V ( x)][ ( x,0)  
ℏ x

2 x 2
]


m im 2  ( x,0)  2  2 ( x,0) i

2iℏ  d exp( 2ℏ )[ ( x,0)   x  2 x 2  ℏ V ( x)]


i i ℏ  2
2
 [1  V ( x)  ] ( x,0)
ℏ 2m x 2

Thus, we have

i ℏ2 2
 ( x,  )  ( x,0)   [  V ( x)] ( x,0) ,
ℏ 2m x 2

which is the same as that derived from the Schrödinger equation. The path integral
formalism leads to the Schrödinger equation for infinitesimal intervals. Since any finite
interval can be thought of a series of successive infinitesimal intervals the equivalence
would still be true.

((Note))

 1/ 2  1/ 2
im 2  2iℏ  im 2 iℏ  2iℏ 
  )   )
2
d exp(  , d exp(   .
2ℏ   m  2ℏ  m  m 

((Mathematica))

41
 Clear@"Global`"D;

Integral@ n_D :=
m η2
IntegrateBη ExpB n
F, 8η, −∞, ∞<F êê
2—ε
SimplifyB , Im B F > 0F &;
m
ε—

K1 = Table@8n, Integral@nD<, 8n, 0, 4<D;

K1 êê TableForm
2π
0
− m
ε—

1 0
2π
2 m 3ê2
J− N
ε—
3 0
2π
3
4 m 5ê2
J− N
ε—

8. Motion of free particle; Feynman path integral

The Lagrangian of the free particle is given by

m 2
L xɺ .
2

Lagrange equation for the classical path;

d L L
( ) ( )0,
dt xɺ x

or

xɺ  a ,

or

x  at  b .

This line passes through (t', x'), (t, x);

42
 x  x'
x1  x '  (t1  t ' ) ,
t  t'

x1

t,x

t',x'

t1

Then we have

x  x'
x1  x ' (t1  t ' ) ,
t  t'

and

1 dx1 2 1 x  x' 2
L(t1 )  m( )  m( ) .
2 dt1 2 t  t'

which is independent of t1. Consequently, we have

t t t
m x  x' 2 m x  x' 2 m ( x  x' ) 2
S cl   L(t1 )dt1 
2 t ' t  t ' 2 t  t ' t '
( ) dt  ( ) dt  ,
2 t  t'
1 1
t'

and

43
 i  m( x  x' ) 2
K ( x, t , x' , t ' )  A exp[ S cl ]  A exp[ ].
ℏ 2ℏi (t  t ' )

((Approach from the classical limit))

To find A, we use the fact that as t  t ' 0 , K must tend to  ( x  x ' ) 

1 ( x  x' ) 2
 ( x  x' )  lim exp[  ]
  0 (2 )1 / 2 2
.
1 ( x  x' ) 2
 lim exp[ ]
  0 2  2 2

where


 ,
2

1 ( x  x' ) 2
f ( x , x ' , )  exp[ ]. (Gaussian distribution).
2  2 2

So we get

2ℏi (t  t ' )
 ,
m

1 m
A  ,
(2 )1 / 2 2ℏi(t  t ' )

or

m  m( x  x' ) 2
K ( x, t , x' , t ' )  exp[ ].
2ℏi (t  t ' ) 2ℏ i ( t  t ' )

Note that

m
Ffreep  article (t  t ' )  .
2ℏi(t  t ' )

((Mathematics))

44
45
((Evaluation of S / ℏ ))
From the above discussion, S can be evaluated as

m ( x )2 m( x ) x mv p h
S   x  x  x ,
2 t 2 t 2 2 2

or

S mv 1 p 1
 x  x  kx .
ℏ 2ℏ 2ℏ 2

where p is the momentum,

46
 p  ℏk .

Suppose that m is the mass of electron and the velocity v is equal to c/137. We make a
S
plot of (radian) as a function of x (cm).
ℏ

((Mathematica))
NIST Physics constant : cgs units
Clear@"Global`∗"D;

rule1 = 9c → 2.99792 × 1010 , — → 1.054571628 10−27 ,

me → 9.10938215 10−28 =;

ê. rule1
me c
K1 =
2—
1.2948 × 1010

K1 ê 137
9.4511 × 107

SêÑ H rad L
1010

108

106

104

100

Dx HcmL
10-8 10-6 10-4 0.01 1 100

S
Fig. (radian) as a function of x (cm), where v = c/137. m is the mass of electron.
ℏ

9. Evaluation of S for the 1D system (example 8-1, Towmsend, 2nd edition)

47
We consider the Young’s double slit;

mx 2 mx 1 1h ℏ
S  x  px  x x.
2( t ) 2 t 2 2 

The phase difference between two paths is evaluated as

S 1 p 1 
 x  kx  x ,
ℏ 2 ℏ 2 

S
If is comparable to , the interference effect can be observed. Such a condition is
ℏ
satisfied when

x   .

((Note))

In classical physics, the phase difference is given by

2
  x .


10. What is the Lagrangian for photon?

((Landau-Lifshitz))
For a particle, we have the Hamilton equations

H H
pɺ   , v  rɺ 
r p

48
In view of the analogy, we can immediately write the corresponding equation for rays:

 
kɺ   , v  rɺ 
r k

In vacuum,   ck , so that kɺ  0 , v  cn ( n is a unit vector along the direction of

propagation); in other words, in vacuum the rays are straight lines, along which the light
travels with velocity c.
Pursuing the analogy, we can establish for geometrical optics a principle analogous to
the principle of least action in mechanics. However, it cannot be written in Hamiltonian
form as   Ldt  0 , since it turns out to be impossible to introduce, for rays, a function
analogous to the Lagrangian of a particle. Since the Lagrangian of a particle is related to
the Hamiltonian H by the equation

H
L  p H ,
p

replacing the Hamiltonian H by the angular frequency  and the momentum by the wave

vector k, we should have to write for the Lagrangian in optics, k    . But this
k
expression is equal to zero, since   ck .But this expression is also clear directly from
the consideration that the propagation of rays is analogous to the motion of particles with
zero mass.
As is well, in the case where the energy is constant, the principle of least action for
particles can also be written in the form of the so-called principle of Maupertuis:

 S    p  dl  0

where the integration extends over the trajectory of the particle between two of its points.
In this expression the momentum is assumed to be a function of the energy and the
coordinates. The analogous principle for rays is called Fermat’s principle. In this case, we
can write by analogy:

   k  dl  0


In vacuum, k  n , and we obtain ( dl  n  dl :
c

  dl  0

which corresponds to rectilinear propagation of the rays.

49
((REFERENCE))
L.D. Landau and E.M. Lifshitz, The Classical Theory of Fields (Pergamon, 1996).

((Note))
From a view-point of the phaser diagram for the interference, photons are one of the
best examples. I will use this example for the explanation of photon propagator. It seems
to me that theorists hesitate to use the word of photon. Why is that? It is hard to find the
i
expression of the action S for photon, even if the path integral is expressed by exp( px) ;
ℏ
plane-wave form. The Lagrangian form of photon is not simple compared to that of
particle (such as electron).

11. Single slit experiment

In order to understand the path integral method, let us go back to the Young’s double
slit experiment. We obtain an interference pattern, independent of whether we use a light
source, or particle (electron) source. This can, of course, be explained by saying that there
is a probability amplitude associated with each path. Note that a path integral approach
offers a road to quantum mechanics for systems that are not readily accessible via
Hamiltonian mechanics (Merzbacher, 1998). In his book, Feynman discussed the
principle of least action for particles (Feynman, 1964).

Is it possible to say intuitively that

L  T  cp ,

where T is the kinetic energy of photon, c is the velocity of light, and p is the
momentum.? The momentum p is expressed by

2 ℏ
p  ℏk  ,


where  is the wave length of photon. Then the action S can be evaluated as

S   Ldt   cpdt  cpt  px

where p is assumed to be constant and x  ct . Then we have

i i
exp( S )  exp( px )  exp(ikx )
ℏ ℏ

The phase of change in wave function for photon is

  k x

where k is the wave number.

50
________________________________________________________________________
Imagine the slit divided into many narrow zones, width y (=  = a/N). Treat each as
a secondary source of light contributing electric field amplitude E to the field at P.

We consider a linear array of N coherent point oscillators, which are each identical,
even to their polarization. For the moment, we consider the oscillators to have no intrinsic
phase difference. The rays shown are all almost parallel, meeting at some very distant
point P. If the spatial extent of the array is comparatively small, the separate wave
amplitudes arriving at P will be essentially equal, having traveled nearly equal distances,
that is

E0
E0 (r1 )  E0 (r2 )  ...  E0 (rN )  E0 (r ) 
N

The sum of the interfering spherical wavelets yields an electric field at P, given by the
real part of

E  Re[ E0 (r )ei ( kr1 t )  E0 (r )ei ( kr2 t )  ...  E0 (r )ei ( krN t ) ]
 Re[ E0 (r )ei ( kr1 t ) [1  e ik ( r2 r1 )  eik ( r3 r1 )  ...  eik ( rN r1 )) ]]

51
((Note))
When the distances r1 and r2 from sources 1 and 2 to the field point P are large compared
with the separation , then these two rays from the sources to the point P are nearly
parallel. The path difference r2 – r1 is essentially equal to  sin.
Here we note that the phase difference between adjacent zone is

a
k (r2  r1 )    k sin   k ( sin  )
N
k (r3  r2 )  
k (r4  r3 )  

k (rN  rN 1 )  

2
where k is the wavenumber, k  . It follows that


k (r2  r1 )  
k (r3  r1 )  2
k (r4  r1 )  3

(rN  r1 )  ( N  1)

Thus the field at the point P may be written as

E  Re[ E0 (r )ei ( kr1 t ) [1  ei  ei 2  ...  e i ( N 1) ]]

We now calculate the complex number given by

Z  1  ei  ei 2  ...  ei ( N 1)

1  eiN

1  ei
e iN / 2 (eiN / 2  e  iN / 2 )
 i / 2 i / 2
e (e  e  i / 2 )
N
sin( )
 ei ( N 1) / 2 2

sin( )
2

If we define D to be the distance from the center of the line of oscillators to the point P,
that is

52
 1 1
kD  ( N  1)k sin   kr1  ( N  1)  kr1
2 2
1
k ( D  r1 )  ( N  1)
2

Then we have the form for E as

N
sin( )
E  Re[ E0 (r )ei ( kD t ) 2 ]  Re[ E~e it ]

sin( )
2

The intensity distribution within the diffraction pattern due to N coherent, identical,
distant point sources in a linear array is equal to

c 0 ~ 2
I S  E
2

N  
sin 2 ( ) sin 2 ( ) sin 2 ( )
I  I0 2 I 2 I 2
2  
0 m 2
sin ( ) 2
sin ( )  
2 2N  
2

in the limit of N→∞, where

2
   
2
sin ( ) 
2N  2N 

c 0
I0  [ E0 (r )]2
2

c 0 c
I m  I0 N 2  [ NE0 (r )]2  0 E0
2

2 2

  N  Nk sin   ka sin 

  k sin 

where a = N. We make a plot of the relative intensity I/Im as a function of .

53
 
sin 2 ( )
I 2
 2
Im 
 
 
2

Note that



sin 2 ( )
   22 d  2
 
2
IêIm
0.06

0.05

0.04

0.03

0.02

0.01

bê2p
-4 -2 0 2 4

The numerator undergoes rapid fluctuations, while the denominator varies relatively
slowly. The combined expression gives rise to a series of sharp principal peaks separated
by small subsidiary maxima. The principal minimum occurs in directions in direction m
such that

54
  ka
 sin   m
2 2
1 
a sin  m  2m  2m  m
k 2

12. Phasor diagram

(i) The system with two paths
The phasor diagram can be used for the calculation of the double slilts (Young)
interference. We consider the sum of the vectors given by OS and ST . The magnitudes
of these vectors is the same. The angle between OS and ST is  (the phase difference).

Fig. Phasor diagram for the double slit.

In this figure, QO  QS  QT  R . SOM  STM   / 2 . Then we have

 
OT  2OM  2OS cos  2 A cos .
2 2

The resultant intensity is proportional to OT ,   2

 
I  OT
2
 4 A 2 cos 2

2
 2 A 2 (1  cos  ) .

Note that the radius R is related to OS (= A) through a relation

55
 
A  2 R sin .
2

When A = 1 (in the present case), we have the intensity I as


I  4 cos 2
2

The intensity has a maximum (I = 4) at   2n and a minimum (I = 0) at

  2 (n  1 / 2) .

(ii) The system with 6 paths.

Fig. The resultant amplitude of N = 6 equally spaced sources with net successive phase
difference φ.  = N φ = 6 φ.

(iii) The system with 36 paths (comparable to single slit)

56
Fig. The resultant amplitude of N = 36 equally spaced sources with net successive
phase difference φ.

(iv) Single slit in the limit of N→∞

We now consider the system with a very large N. We may imagine dividing the slit
into N narrow strips. In the limit of large N, there is an infinite number of infinitesimally
narrow strips. Then the curve trail of phasors become an arc of a circle, with arc length
equal to the length E0. The center C of this arc is found by constructing perpendiculars at
O and T.

The radius of arc is given by

57
 E0  R  R( N ) .

in the limit of large N, where R is the side of the isosceles triangular lattice with the
vertex angle φ, and  is given by

  N  ka sin  ,

with the value  being kept constant. Then the amplitude Ep of the resultant electric field
at P is equal to the chord OT , which is equal to


sin
 E0  2 .
EP  2 R sin 2 sin  E0
2  2 
2
Then the intensity I for the single slits with finite width a is given by


sin
I  Im ( 2 )2 .

2

where Im is the intensity in the straight-ahead direction where 

a
The phase difference φ is given by   ka sin   2 sin   2p sin  . We make a

plot of I/Im as a function of , where p = a/ is changed as a parameter.

Fig. The relative intensity in single-slit diffraction for various values of the ratio p =
a/. The wider the slit is the narrower is the central diffraction maximum.

13. Gravity: Feynman path integral

58
((Calculation from the classical limit))

m 2
L xɺ  mgx ,
2

d L L
( )  ( ),
dt xɺ x

ɺxɺ   g ,

g 2
x1   t1  At1  B ,
2

Initial conditions:

x1 = x and t1= t,

g
x   t 2  At  B .
2

and x1  x' and t1  t '

g 2
x'   t '  At ' B ,
2

A and B are determined from the above two equations.

 (t 't1 )[ g (t  t ' )(t  t1 )  2 x]  2(t  t1 ) x '

x1  ,
2(t  t ' )

dx1 g (t  t ' )(t  t '2t1 )  2( x  x ' )

xɺ1   .
dt1 2(t  t ' )

Then we get the expression of the Lagrangian

m 2
L( x1 , xɺ1 , t1 )  xɺ1  mgx1 .
2

The Hamilton’s principle function is

59
 t
S cl ( x, t , x' , t ' )   L( x1, xɺ1 , t1 )dt1
t'

m[ g 2 (t  t ' ) 4  12( x  x ' ) 2  12 g (t  t ' )2 ( x  x' )


24(t  t ' )

i
K ( x, t ; x' , t ' )  A exp[ Scl ( x, t : x ' , t ' )] .
ℏ

((Classical limit))

When x  x' and t  t 'T , we have

1
S cl ( x, t : x' , t ' )   mg 2T 3  x' gmT ,
24

and

i ix' mgT
K ( x, t ; x' , t ' )  A exp( mg 2T 3  )
24ℏ ℏ
ix' mgT
 A1 exp( )
ℏ

where

i
A1  A exp( mg 2T 3 ) .
24ℏ

((Mathematica))

60
 −1
Clear@"Global`∗"D; eq1 = x g t2 + A t + B;
2
−1
eq2 = x0 g t02 + A t0 + B;
2

rule1 = Solve@8eq1, eq2<, 8 A, B<D êê Flatten;

−1
x1 = g t12 + A t1 + B ê. rule1 êê FullSimplify
2
− Ht0 − t1L Hg Ht − t0L Ht − t1L + 2 xL + 2 Ht − t1L x0
2 Ht − t0L

v1 = D@x1, t1D êê Simplify

g Ht − t0L Ht + t0 − 2 t1L + 2 Hx − x0L
2 Ht − t0L

v12 − m g x1 êê FullSimplify
m
L1 =
2
1
m IHg Ht − t0L Ht + t0 − 2 t1L + 2 Hx − x0LL2 −
8 Ht − t0L2
4 g Ht − t0L H− Ht0 − t1L Hg Ht − t0L Ht − t1L + 2 xL + 2 Ht − t1L x0LM

K1 = ‡ L1 t1 êê FullSimplify
t

t0

m Ig2 Ht − t0L4 − 12 Hx − x0L2 + 12 g Ht − t0L2 Hx + x0LM

−
24 Ht − t0L

rule1 = 8x → x0, t → t0 + T<; K2 = K1 ê. rule1 êê Simplify

1
− g m T Ig T2 + 24 x0M
24

14. Simple harmonics: Feynman path integral

((Classical limit))

m 2 1
xɺ  m 0 x 2 ,
2
L
2 2

d L L
( ) ,
dt xɺ x

ɺxɺ  0 2 x ,

61
 x1  A cos(0t1 )  B sin(0t1 ) .

Initial conditions:

x1 = x and t1= t,

x  A cos 0 t  Bsin  0 t ,

and

x1  x' and t1  t ' ,

x'  A cos 0t ' B sin 0t ' .

A and B are determined from the above two equations.

x' sin[0 (t  t1 )]  x sin[0 (t 't1 )]

x1  ,
sin[0 (t  t ' )]

dx1  x'0 cos[0 (t  t1 )]  x0 cos[0 (t 't1 )]

xɺ1   .
dt1 sin[0 (t  t ' )]

Then we get the expression of the Lagrangian

m 2 1
L( x1 , xɺ1, t1 )  xɺ1  m0 x1
2 2

2 2
m0
2
 [2 xx ' cos[0 (t  t '2t1 )
2 sin 2 [0 (t  t ' )]
 x'2 cos[ 20 (t  t1 )]  x 2 cos[0 (t 't1 )])

The Hamilton’s principle function is

t
S cl ( x, t , x' , t ' )   L( x1 , xɺ1 , t1 )dt1
t'

m0
 [( x 2  x '2 ) cos(0 (t  t ' ))  2 xx ' ]
2 sin(0 (t  t ' )

i
K ( x, t ; x' , t ' )  A exp[ Scl ( x, t : x ' , t ' )] ,
ℏ

or

62
 im0
K ( x, t ; x' , t ' )  A exp[ {( x 2  x'2 ) cos(0 (t  t ' ))  2 xx'}] .
2ℏ sin[0 (t  t ' )]

In the limit of t  t ' 0 , we have

im0
K ( x, t , x' , t ' )  A exp[ ( x  x' ) 2 ] .
2ℏ sin{0 (t  t ' )}

To find A, we use the fact that as t  t ' 0 , K must tend to  ( x  x ' ) ,

1 ( x  x' ) 2
 ( x  x' )  lim exp[  ]
  0 (2 )1 / 2 2
.
1 ( x  x' ) 2
 lim exp[ ]
  0 2  2 2

where


 ,
2

1 ( x  x' ) 2
f ( x , x ' , )  exp[ ] (Gaussian distribution).
2  2 2

In other words

1 1 m0
A ,  .
(2 )1/ 2 2
2iℏ sin{0 (t  t ' )}

So we get

2iℏ sin{0 (t  t ' )} 1 m0

 , A 
m0 (2 )1 / 2 2ℏi sin{0 (t  t ' )}

or

m0
K ( x, t , x' , t ' ) 
2ℏi sin{0 (t  t ' )}
im0
 exp[ [( x 2  x'2 ) cos{0 (t  t ' )}  2 xx ' ]]
2ℏ sin[0 (t  t ' )]

63
Note that

m0
K ( x  0, t , x'  0, t ' )   F (t  t ' )
2ℏi sin{0 (t  t ' )}

and

i
K ( x, t , x' , t ' )  F (t  t ' ) exp[ S cl ( x, t; x' t ' )] .
ℏ

((Mathematica))
Clear@"Global`∗"D;
expr_∗ := expr ê. 8Complex@a_, b_D Complex@a, − b D<

seq1 = x A Cos@ω0 tD + B Sin@ω0 tD;

seq2 = x0 == A Cos@ω0 t0D + B Sin@ω0 t0D;

srule1 = Solve@8seq1, seq2<, 8 A, B<D êê Simplify êê Flatten;

x1 = A Cos@t1 ω0 D + B Sin@t1 ω0 D ê. srule1 êê Simplify;

v1 = D@x1, t1D êê Simplify;

ω0 2 x12 êê Simplify;
m m
L1 = v12 −
2 2

‡ L1 t1 êê FullSimplify
t

t0
1
m ω0 I− 2 x x0 + Ix2 + x02 M Cos@Ht − t0L ω0DM Csc@Ht − t0L ω0D
2

If the initial state of a harmonic oscillator is given by the displaced ground state wave
function

m 0
 (x,0)  exp[ (x  x 0 ) 2].
2ℏ

When

  x ,

with

m0
 .
ℏ

Then we have

64
 1 1
 ( ,0)  exp[  (  0 ) 2 ] ,
 1/ 4
2

and

 ( , t )   K ( , t; ' ,0) ( ' ,0)d '


1 i0t sin(0t ) i0t

 exp(  ) exp[  e {( 2  0 ) cot(0t )  i 2
2

 1/ 4
2 2
1
 20 }]
sin(0t )
1 i0t 1
 exp(  ) exp[  e i0t {( 2  0 ) cos(0t )  i 2 sin(0t )
2

 1/ 4
2 2
 20 }]
i0t
1 i0t 1 2 e  e i0t
 exp(  ) exp[  e i0t {( 2e i0t  0 ( )
 1/ 4
2 2 2
 20 }]
1 i0t 1 2 1 2
 exp[   {  20e i0t  0 (1  e 2i0t )}]
 1/ 4
2 2 2

or

1 i0t 1 2
 ( , t )  exp[   {  20 (cos 0t  i sin 0t )
 1/ 4
2 2
1 2
 0 (1  cos 20t  i sin 20t )}]
2
1 1 t 1 2
 1 / 4 exp[  (  0 cos 0t ) 2  i ( 0  0 sin 0t  0 sin 20t )]
 2 2 4

Finally, we have

1
| ( , t ) |2  exp[ (  0 cos(0t )) 2 ]
 1/ 2

((Mathematica))

65
 Clear@ "Global`∗"D ;
exp_ ∗ := exp ê. 8 Complex@ re_, im_D Complex@ re, − im D< ;

KSH@ ξ_, t_, ξ1_ D := ExpB II ξ 2 + ξ 12 M Cos@ ω0 tD − 2 ξ ξ 1MF ;

1
2π Sin@ ω0 tD 2 Sin@ω 0 tD

H ξ − ξ0 L2
ϕ 0@ ξ _D := π−1ê4 ExpB− F;
2

KSH@ξ , t, ξ1 D ϕ0 @ξ1 D ξ1 êê FullSimplify@ , 8 Im @ Cot@t ω0DD > − 1, ω 0 t > 0<D &

∞
f1 = ‡
−∞

Jξ 2 +ξ0 2 N Cot@t ω0 D+ ξ Hξ+2 ξ0 Csc@t ω0 DL

−
2 H +Cot@t ω0 DL − Csc@t ω 0D
π1ê4 1− Cot@t ω0 D

Amp1 = f1∗ f1 êê FullSimplify

−Hξ−ξ0 Cos@t ω0 DL2

π
rule1 = 8ω 0 → 1, ξ0 → 1<; H1 = Amp1 ê . rule1;

Plot@Evaluate@Table@H1, 8t, 0, 20, 2<D, 8 ξ , − 3, 3<D ,

PlotStyle → Table@8Thick, Hue@ 0.1 iD< , 8 i, 0, 10<D , AxesLabel → 8 "ξ ", "Amplitude"<D
Amplitude

0.5

0.4

0.3

0.2

0.1

x
-3 -2 -1 1 2 3

15. Gaussian wave packet propagation (quantum mechanics)

x  (t )  x Uˆ (t , t ' )  (t ' )   dx ' x Uˆ (t , t ' ) x' x'  (t ' ) ,

i
K ( x, t ; x ' , t ' )  x Uˆ (t , t ' ) x'  x exp[ Hˆ (t  t ' )) x ' ,
ℏ

or

x  (t )   dx' K ( x, t ; x ' , t ' ) x'  (t ' ) .

K(x, t; x’, t’) is referred to the propagator (kernel)

For the free particle, the propagator is given by

66
 m im( x  x' ) 2
K ( x, t ; x ' , t ' )  exp[ ].
2iℏ(t  t ' ) 2ℏ ( t  t ' )

Let’s give a proof for this in the momentum space.

Hˆ is the Hamiltonian of the free particle.

1 ikx
xk  e ,
2

Hˆ k  Ek k ,

with

ℏ 2k2 Ek
Ek  , k  ,
2m ℏ

i
K ( x, t ; x ' , t ' )  x exp[ Hˆ (t  t ' )) x'
ℏ
i
  dk x k k exp[ Hˆ (t  t ' )) x'
ℏ
iℏk 2
  dk x k k exp[ (t  t ' )) x'
2m
1 iℏk 2
  dk exp[ik ( x  x' )  (t  t ' )]
2 2m

Note that

i ℏk 2 iℏ (t  t ' ) m( x  x' ) 2 m( x  x' ) 2

exp[ik ( x  x' )  (t  t ' )]  exp[ {k  }  ],
2m 2m ℏ (t  t ' ) 2iℏ (t  t ' )

and



 dk exp(ik )
2
,

i

or

m im( x  x' ) 2
K ( x, t ; x ' , t ' )  exp[ ].
2iℏ(t  t ' ) 2ℏ ( t  t ' )

67
 ((Quantum mechanical treatment))

Probability amplitude that a particle initially at x’ propagates to x in the interval t-t’. This
expression is generalized to that for the three dimensions.

m im | r  r ' |2
K (r , t; r ' , t ' )  [ ]3 / 2 exp[ ].
2iℏ (t  t ' ) 2ℏ(t  t ' )

We now consider the wave function:

1 x2
x  (t  0)  exp(ik 0 x  ).
4 x
2
2  x

where the probability

2 1 x2
P  x  (t  0)  exp( ),
2  x 2 x
2

has the form of Gaussian distribution with the standard deviation  x . The Fourier
transform is given by

k  (t  0)   dx k x x  (t  0)
1 1 x2
2 
 dx exp(ikx) exp(ik0 x  )
4 x
2
2  x
1 1
 2 x  exp[ x (k  k0 ) 2 ]
2

2  x 2
1/ 4
2
   x exp[ x 2 (k  k0 ) 2 ]
 

or

2 2
k  (t  0)   x exp[2 x 2 (k  k0 ) 2 ]

1 1 ( k  k0 ) 2
 exp[ ].
2  1 
2
   1 
2 
 2 x   2 x 
1 ( k  k0 ) 2
 exp[ ]
2  k 2 k
2

68
  1 
which is the Gaussian distribution with the standard deviation  k    .
 2 x 
Then we get the wave function at time t,

x  (t )   dx' K ( x, t ; x' ,0) x'  (0)

1 m x'2 im( x  x ' ) 2
2iℏt 
 dx ' exp[ik 0 x '   ]
4 x
2
2  x 2ℏt

mx ( x  4ik0 x )  2ik 0 tℏ x

2 2 2
1 m 1
 2  exp[ ]
2  x 2iℏt 1 2im 4m x  2itℏ
2

x 2
tℏ
tℏ 2
x( x  4ik 0 x )  2ik0 x
2 2
1 1 m
 exp[ ]
2  x itℏ itℏ
4 x (1 
2
1 )
2m x
2
2m x
2

tℏ 2 itℏ
{x( x  4ik 0 x )  2ik 0  x }(1 
2 2
)
2m x
2
1 1 m
 exp[ ]
2  x itℏ itℏ itℏ
4 x (1  )(1 
2
1 )
2m x 2m x
2 2
2m x
2

ℏk 0 t 2
(x 
)
m(8k0 mx x  t ( x 2  4k0  x )ℏ )
4 2 4
1 1 m
 exp[  i ]
2  x itℏ t 2ℏ 2 t 2ℏ2
4 x (1  8m  x (1 
2 2 4
1 ) )
4m 2 x 4m 2 x
4 4
2m x
2

Since

1 1

1
itℏ t 2ℏ 2
1
2m x
2
4m 2 x
4

the probability is obtained as

ℏk0t 2
(x  )
2 1 m
x  (t )  exp[ ].
t 2ℏ 2 t 2ℏ 2
2 x (1 
2
2  x 1 )
4m 2 x
4
4m 2 x
4

69
which has the form of Gaussian distribution with the standard deviation

t 2ℏ 2
x 1 .
4m 2 x
4

and has a peak at

ℏk 0 t
x  .
m

The Fourier transform:

1
k  (t )   dx k x x  (t )   dxe  ikx x  (t )
2
tℏ 2
x( x  4ik0 x )  2ik0 x
2 2
1 1 1 m

2 2  x itℏ  dx exp[ikx  itℏ
]
4 x (1 
2
1 )
2m x
2
2m x
2

1 ( k  k0 ) 2 ik 2tℏ
 exp[ 2
 ]
2
1  1  2m
4 
2 x  2 x 

where

tℏ 2
x( x  4ik0 x )  2ik0 x
2 2
1 1 m
x  (t )  exp[ ].
2  x itℏ itℏ
4 x (1 
2
1 )
2m x
2
2m x
2

Then the probability is obtained as

2 1 ( k  k0 ) 2
k  (t )  exp[ ].
2  k 2 k
2

where

1
k  .
2 x

Therefore

70
 2 2
k  (t )  k  (t  0) .

H»yHx,tL»L2

2.0

1.0

0.5

0.2

0.1

x
0 1 2 3 4 5

2
Fig. Plot of x  (t ) as a function of x where the time t is changed as a parameter.

In summary, we have

ℏk0t 2
(x  )
2 1 m
x  (t )  exp[ ].
t 2ℏ 2 t 2ℏ 2
2 x (1 
2
2  x 1  )
4m 2 x
4
4m 2 x
4

and

2 1 ( k  k0 ) 2
k  (t )  exp[ ].
2  k 2 k
2

16. Wave packet for simple harmonics (quamtum mechanics)

((L.I. Schiff p.67-68))

71
 i
x  (t )  x exp( Hˆ t )  (t  0)
ℏ
i
  x exp( Hˆ t ) x' x'  (t  0) dx'
ℏ

We define the kernel K ( x, x' , t ) as

i
K ( x, x ' , t )  x exp( Hˆ t ) x '
ℏ
i
  x n exp( Ent ) n x '
n ℏ
i
  exp( Ent ) n ( x) n* ( x' )
n ℏ

Note that

  x ,

with

m0
 .
ℏ

Then we have

i
 ( x, t )   exp( Ent )  dx' n* ( x ' ) ( x' ) n ( x) .
n ℏ

We assume that

 1/ 2 1
 ( x)  exp[  2 ( x  a )2 ] ,
 1/ 4
2

or

1 1
 n ( )   n  xn  n ( x) ,
 

or

1 1
 ( )  exp[ (   0 ) 2 ] ,
 1/ 4
2

72
with

0   x0 .

We need to calculate the integral defined by

I   dx ' n* ( x ' ) ( x' )

 1/ 2 1
  dx ' n* ( x ' ) exp[  2 ( x ' a )2 ]
 1/ 4
2
d 1 / 2 *  1/ 2
1
   n ( ) 1 / 4 exp[ (   0 )2 ]
  2
1 1
  d ( ) exp[ 2 (   ) ]
* 2

 1/ 4 n 0

Here
2
1 
 
 n ( )  (  2 n!) e n 2 2
H n ( ) .

Then we get

1 
1
2 1
I (  2 n!)  d exp(  )Hn ( )exp[  (  0 ) ]) .
n 2 2

 1 /4
2 2

Here we use the generating function:


sn
exp(2s  s )   Hn ( ).
2

n 0 n!

Note that

 
2 1 
sn 2 1
 d exp(2s  s ) exp(
2
) exp[ (  0 ) 2 ]    d H n ( ) exp( ) exp[ (   0 )2 ]

2 2 n 0  n! 2 2
 
sn 1 2
   d H n ( ) exp[( 2  0   0 )]
n 0  n! 2

The left-hand side is

73
 
1 2  2

d exp(2s  s ) exp[(   0  2 0 )]   exp(s0  40 )

2 2 1/ 2

02 s n 0
 n
  1 / 2 exp( )
4 n0 n!

Thus we have

0 2 s n 0 n
 
sn 1
 1/ 2 exp( )    d Hn ( )exp[(2  0  0 2 )] .
4 n 0 n! n0  n! 2

or

0 2 1
 1/ 2 exp( )0 n   dH () exp[(  0  0 2 )] .
2
n
4 
2
Then


1
0 2
I  (  2 n!) exp(  )0 ,
n 2 n

and

i 
1
0 2
 (x,t)   exp(  En t)(2 n n!) 2 exp(  )0  n (x) .
n

n ℏ 4

Since

1
En  ℏ 0 (n  ) ,
2

or

i i
exp(  En t)  exp(   0 t  in  0 t) ,
ℏ 2

and

1
 ( , t )   ( x, t ) ,


we get

74
  
1
0 2 i 0 t n i
 (,t)   (2n n!) 2 exp(  )(0 e ) exp(   0 t) n () ,
n 0 4 2

or

 2
1 
0 2 i 
 ( , t ) 
 1/4
 (2 n !)
n 0
n 1
exp(
4
)( 0e  i 0 t n
) exp( 0t )e 2 H n ( )
2
1 i  2  1  0 e  i 0 t n
0 2
 exp(  0t   ) ( ) H n ( )
 1/4 4 2 2 n 0 n ! 2

Using the generating function

 0t
1 0 e i

1 n t  t
 n! ( 2 )n Hn ( )  exp[ 0 2 e2i 0  0e i 0  ],
4
n0

we have the final form

1 i 0 2
1 2  0t 2  0t
 ( , t )  exp(  0t   0 e  2i   0e  i )

 1/ 4
4 2 2 4
1  2
 2
i 1 2
 1 / 4 exp[ 0   0t   0 (cos 20t  i sin 20t )
 4 2 2 4
 0 (cos 0t  i sin 0t )]

OR

2 1 0 2 1
 ( ,t)  1/ 2 exp[   2  0 2 cos2 0t  20  cos 0 t]
 2 2

or

2 1
 ( , t )  exp[(  0 cos 0t )2 ]
 1/ 2

2
 ( ,t) represents a wave packet that oscillates without change of shape about  = 0
with amplitude 0 and angular frequency 0.

17. Mathematica

75
 2
,t

0.6

0.4

0.2

4 2 2 4

2 1
Fig. The time dependence of  ( , t )  exp[(  0 cos 0t ) 2 ] , where  0 =1.
 1/ 2

T  2 / 0 . The peak shifts from  = 0 at t = 0 to  = 0 at t = T/4,  =  0 at t =

T/2,  = -  0 at t = 3T/4, and  = 0 at t = T.

76
77
18. Neutron interferometry
A neutron beam is split into two beams by an interferometer. The relative phase of the
two beams is varied by rotating the interferometer around the AC of the incident beam.
(R. Colella, A.W. Overhauser, and S.A. Werner [Phys. Rev. Lett. 34, 1472 (1975)].
Suppose that the interferometry initially lies in a horizontal plane so that there are no
gravitational effects. We then rotate the plane formed by the two paths by angle  about
the segment AC. The segment BD is now higher than the segment AC by l2sin.

78
Fig. Experimental cobfiguration for neutron interefernce due to gravity. There
are two paths; path A-B-D and path A-C-D.  is the rotation angle from
the z axis in the z-x plane.

79
Fig. Dependence of gravity-induced phase on angle of rotation . From R.
Colella, A. W. Overhauser, and S. A. Werner, Phys.. Rev. Lett. 34 ( 1975)
1472.

((Note))
The action is obtained as

m[ g 2 (t  t ')4  12( x  x ') 2  12 g (t  t ') 2 ( x  x ')]

S cl ( x, t , x ', t ')  
24(t  t ')

l1
where x is the height Along the path BD, t  t '  T  , and x  x '  x0  l1 sin  . Along
v
l1
the path AC, t  t '  T  , and x  x '  x0  0
v

m( g 2T 4  12 gT 2 2 x0 )
S cl ( x0 , T )  
24T
m( g T  24Tgx0 )
2 3

24

From the previous discussion we have the propagator for the gravity

80
 i
K ( x, t ; x0 , t0 )  A exp( Scl )
ℏ
i ix mgT
 exp(  mg 2T 3  0 )
24ℏ ℏ
ix mgT
 A1 exp(  0 )
ℏ

For the path ABD and path ACD, we have

i i i
  exp[ S ( ABD)]  exp[ S ( ACD )]  exp[ S ( ACD )]{1  exp(i )} .
ℏ ℏ ℏ

The phase difference between the path ABD and the path ACD is given by

S ( ABD )  S ( ACD ) 1 1 m 2 gl1l2

    mgl2T sin    sin  ,
ℏ ℏ ℏ p

or

1 m 2 gl1l2 m 2 gl1l2 
   sin    sin    sin  .
2ℏ 2 ℏ 4 ℏ 2


or

1 m 2 gl1l2 mgl1l2
   sin    sin    sin  ,
2ℏ mv 2ℏv

where p is the momentum

p  mv .

The time T is related to l1 as

l1 ml1 ml1
T   .
v p 2ℏ


The intensity for the interference is

81
 2
  1  e i 
2

1
 4cos 2 (  ) .
2
 sin 
 4cos 2 ( )
2

Fig. The phasor diagram. The intensity corresponds to the length OT , where
OS  ST  1.

Note that

mgl1l2
  ,
2ℏv

where m is the mass of neutron. We consider the thermal neutron at the temperature
1 3
T  300 K. The energy of neutron is given by E  mv 2  k BT . The thermal energy is
2 2

E  38.778 meV.

The average velocity v is evaluated as

82
 3k BT
v  2.72374 103 m/s.
m

The wavelength is

2 ℏ
  1.452 Ǻ.
mv

So that, we have

mg
 2.85931 (1/cm2)
2 ℏv

The intensity I is evaluated as

mgl1l2
I  4 sin 2 ( cos  )  4 sin 2 (1.4297l1l2 cos  )
4 ℏv

We assume that the area l1l2  6 cm 2 . We make a plot of the intensity I as a function of
the rotation angle  (radian).

Fig. Plot of the intensity (neutron interference due to gravity) as a function of

rotation angle  (rad).

((Mathematica)) We use the cgs units for the calculation.

83
84
19. Quantum-mechanical interference to detect a potential difference

A low-intensity beam of charged particles, each with charge q, is split into two parts.
each part then enters a very long metallic tube shown above. Suppose that the length of
the wave packet for each of the particles is sufficiently smaller than the length of the tube
so that for a certain time interval, say from t0 to t, the wave packet for the particle is
definitely within the tubes. During this time interval, a constant electric potential V1 is
applied to the upper tube and a constant electric potential V2 is applied to the lower tube.
The rest of the time there is no voltage applied to the tubes. Here we consider how the
interference pattern depends on the voltages V1 and V2.

Without the applied potentials, the amplitude to arrive at a particular point on the
detecting screen is

85
   1  2 .

The intnsity is proportional to

2 2 2
I 0   1   2   1   2  ( 1  2   1 2 ) .
* *

With the potential, the Lagrangian in the path is modified as

L0  L  L0  qV .

Thus the wave functions are modified as

t t
i i
   1 exp[ 
ℏ t0
( qV1 )dt ]   2 exp[  ( qV12 )dt ]
ℏ t0
iqV1 iqV2
  1 exp[  (t  t0 )]   2 exp[ (t  t 0 )]
ℏ ℏ
iqV1 iqV2
  1 exp[  t ]   2 exp[ t ]
ℏ ℏ

where t  t  t0 . The intensity of the screen is proportional to

2
I 
2
iqV1 iqV2
  1 exp[  t ]   2 exp[  t ]
ℏ ℏ
2
iq(V1  V2 )
  1   2 exp[ t ]
ℏ

We assume that

q (V1  V2 )
 t .
ℏ

Then we have
2
I   1  2 ei   1   2  ( 1* 2 ei  1 2*e i )
2 2

When  1   2   0 , we have

2 2 
I  2  0 (1  cos  )  4  0 cos 2
2

86
The intensity depends on the phase; I becomes maximum when   2n and minimum
at

1
  2(n  ) .
2

((Note)) Method with the use of gauge transformation

The proof for the expression can also be given using the concept of the Gauge
trasnformation. The vector potential A and scalar potential  are related to the magnetic
field B and electric field E by

B  A,

1 A
E   ,
c t

The gauge transformation is defined by

A'  A   ,

1 
'   ,
c t

where  is an arbitrary function. The new wave function is related to the old wave
function through

iq
 '(r )  exp(  ) (r ) .
ℏc

Suppose that  ' 0 . Then  ' (r )   0 (r ) is the wave function of free particle. Then we
get

1 
 ,   c  dt .
c t

The wave function  (r ) is given by

87
 iq
 (r )  exp(  ) 0 (r )
ℏc
iq
 exp( c  dt ) 0 (r )
ℏc
iq
 exp(  dtV ) 0 (r )
ℏ

When V is the electric potential and is independent of time t, we have

iq
 (r )  exp[ V (t  t0 )] 0 (r )
ℏ

This expression is exactly the same as that derived from the Feynman path integral.

20. Quantization of magnetic flux and Aharonov-Bohm effect

The classical Lagrangian L is defined by

1 2 q
L mv  q  v  A .
2 c

in the presence of a magnetic field. In the absence of the scalar potential (   0 ), we get

1 2 q ( 0) e
Lcl  mv  v  A  Lc  v  A ,
2 c c

where the charge q = -e (e>0), A is the vector potential. The corresponding change in the
action of some definite path segment going from (rn 1, tn 1 ) to (rn1, tn ) is then given by

t
e n  dr 
c t n1  dt 
S ( 0) (n, n  1)  S (0 ) (n, n  1)  dt    A ,

This integral can be written as

t r
e n  dr  e n
c t n1  dt 
dt    A   A  dr ,
c rn 1

where dr is the differential line element along the path segment.

Now we consider the Aharonov-Bohm (AB) effect. This effect can be usually
explained in terms of the gauge transformation. Here instead, we discuss the effect using
the Feynman’s path integral. In the best-known version, electrons are aimed so as to pass
through two regions that are free of electromagnetic field, but which are separated from

88
each other by a long cylindrical solenoid (which contains magnetic field line), arriving at
a detector screen behind. At no stage do the electrons encounter any non-zero field B.

Fig. Schematic diagram of the Aharonov-Bohm experiment. Electron beams are split
into two paths that go to either a collection of lines of magnetic flux (achieved by
means of a long solenoid). The beams are brought together at a screen, and the
resulting quantum interference pattern depends upon the magnetic flux strength-
despite the fact that the electrons only encounter a zero magnetic field. Path
denoted by red (counterclockwise). Path denoted by blue (clockwise)

89
 Reflector

C1

Incident Electron beams

P slit -1 Q Screen
slit -2
B
out of page

C2

Reflector
Fig. Schematic diagram of the Aharonov-Bohm experiment. Incident electron beams
go into the two narrow slits (one beam denoted by blue arrow, and the other beam
denoted by red arrow). The diffraction pattern is observed on the screen. The
reflector plays a role of mirror for the optical experiment. The path1: slit-1 – C1 –
S. The path 2: slit-2 – C2 – S.

Let  1B be the wave function when only slit 1 is open.

ie
ℏc Path1
 1, B (r )   1, 0 (r ) exp[ dr  A(r )] , (1)

The line integral runs from the source through slit 1 to r (screen) through C1. Similarly,
for the wave function when only slit 2 is open, we have

ie
ℏc Path 2
 1, B (r )   2, 0 (r ) exp[ dr  A(r )] , (2)

The line integral runs from the source through slit 2 to r (screen) through C2.
Superimposing Eqs.(1) and (2), we obtain

ie ie
 B (r )   1, 0 (r ) exp[
ℏc Path1
dr  A(r )]   2,0 (r ) exp[ 
cℏ Path 2
dr  A(r )] .

90
The relative phase of the two terms is

Path1
dr  A(r )  
Path 2
dr  A(r )   dr  A(r )   (  A)  da ,

by using the Stokes’ theorem, where the closed path consists of path1 and path2 along the
same direction. The relative phase now can be expressed in terms of the flux of the
magnetic field through the closed path,

e e e e
 
cℏ  A  dr   (  A)  da   B  da   .
cℏ cℏ cℏ

where the magnetic field B is given by

B   A.

The final form is obtained as

ie
ℏc Path 2
 B (r )  exp[ dr  A(r )][ 1, 0 (r ) exp( i )   2, 0 (r )] ,

and  is the magnetic flux inside the loop. It is required that

  2n .

Then we get the quantization of the magnetic flux,

2cℏ
n  n ,
e

where n is a positive integer, n = 0, 1,2,….. Note that

2cℏ
=4.1356675 10-7 Gauss cm2.
e

which is equal to 2  0 , where  0 is the magnetic quantum flux,

2cℏ
0  = 2.067833758(46) x 10-7 Gauss cm2. (NIST)
2e

We note that

e  e 0 
   
cℏ  0 cℏ 0

91
The intensity is

I  I 0 (1  ei )(1  e  i )

 2 I 0 [1  cos( )]
  
 4cos 2  
 2 
  
 4cos 2   
 0 

I I0
3

1 1 2 3 4 5
x 0

Suppose that the area for the region of magnetic field is A  1 mm 2 =10-2 cm2.

 102 B 105 B
 
 0 2.06783 107 2.06783

If B  0.3G ,  /  0  1.45  104 .

If B  1 mG ,  /  0  48 .
If B  0.01 mG ,  /  0  0.48 .

((Note))

Equivalence between Aharonov Bohm effect and Feynman path integral

92
Lagrangian:

1 2 q
L mv  q  v  A .
2 c

Canonical momentum:

L q
P  mv  A .
v c

Vector potential and scalar potential:

B  A,

1 A
E   ,
c t

The gauge transformation is defined by

A'  A   ,

1 
'    ,
c t

where  is an arbitrary function. The new wave function of free particle (A’=0)  '(r ) is
related to the original wave function  (r ) through

iq
 '(r )  exp(  ) (r ) .
ℏc

For the Aharonov-Bohm effect, we assume that

A '  0  A   ,     A  dr .

So that the original wavefunction  (r ) is

iq
 (r )  exp(  ) '(r )
ℏc
iq
 exp(  ) '(r )
ℏc
iq
  '(r ) exp(  A  dr )
ℏc

93
with

    A  dr ,

 '(r )  exp(ik  r )  exp(i  k  dr ) . (free-particle wave function)

Thus, we have the original wavefunction as

iq
 (r )  exp(i  k  dr ) exp(
ℏc 
A  dr )

i iq
 exp(  ℏk  dr   A  dr )
ℏ ℏc
i q
 exp[ (  ℏk  dr   A  dr )]
ℏ c
i q
 exp[  (p  A )  dr
ℏ c
i
 exp[ S ]
ℏ

The integral is over a certain path in the 3D real space. It is found that that S is the action and
given by

q
S   P  dr   (p  A)  dr , (Feynman path integral)
c

where

L q
P  mv  A .
v c

In conclusion, we show that the Aharonov-Bohm (AB) effect can be explained by the Feynman
path integral. In other words, the AB effect is equivalent to the Feynman path integral.

21. Example-1: Feynman path integral

We consider the Gaussian position-space wave packet at t = 0, which is given by
1 x2
x  (t  0)  exp( 2 ) (Gaussian wave packet at t = 0).
2  2
The Gaussian position-space wave packet evolves in time as

94
 x  (t   dx0 K ( x, t ; x0 ,0) x0  (0)
im( x  x0 ) 2
2
1 m x0
2iℏt 
 dx exp[   ] (1)
2 2
0
2  2ℏt
1 1 x2
 exp[ ]
2 iℏt iℏt
2(  )
2
2 
m m

where

m 1/ 2 im( x  x0 ) 2
K ( x, t; x0 , t0  0)  ( ) exp[ ] (free propagator) (2)
2iℏt 2ℏt

(Note) You need to show all the procedures to get the final form of x  (t .
(a) Prove the expression for x  (t given by Eq.(1).
2
(b) Evaluate the probability given by x  (t for finding the wave packet at the
position x and time t.

(a)
The Gaussian wave packet:

1 x2
x  (t  0)  exp( 2 ) . (Gaussian)
2  2

The free propagator:

m 1/ 2 im ( x  x0 ) 2
K ( x, t ; x0 , t0  0)  ( ) exp[ ].
2iℏt 2ℏt

Then we have

x  (t   dx0 K ( x, t; x0 ,0) x0  (0)

1/ 2
im( x  x0 )2
2
1  m  x0
  
2   2iℏt   dx0 exp[ 2 2  2ℏt ]
Here we have

95
 2 2 2
x im( x  x0 )2 x im( x 2  2 x0 x  x0 )
 02   02 
2 2ℏt 2 2ℏt
1 im 2 imx imx 2
 ( 2  ) x0  ( ) x0 
2 2ℏt ℏt 2ℏt
 ax0  bx0  c
2

b 2 b 2  4ac
 a( x0  ) 
2a 4a

where

1
im imx imx 2
a 2  , b , c .
2 2ℏt ℏt 2ℏt

Then we get the integral

 
b 2 b2
 dx0 exp[ax0  bx0  c]  dx0 exp[a( x0  2a )  c  4a
2



b2 b
 exp(c  )  dx0 exp[a ( x0  )2
4a   2a
 b2
 exp(c  )
a 4a

1
Note that when Re( a )    0 , the above integral can be calculated as
2 2

 
b 2 1 
dx0 exp[a( x0  2a )   dy exp(  y )
2
,
a 
a

b
with the replacement of variable as y   a ( x0  ) and dy  dx0  a . Thus we have
2a

1 m  b2
x  (t  exp( c  )
2  2iℏt a 4a
2
 imx 
 
1 m  imx 2
 ℏt 
 exp[  ]
2 2iℏt 1 im 2 2ℏt 4( 1  im )

2 2ℏt 2 2 2ℏt

or

96
 2
 imx 
 
1 m imx 2
 ℏt 
x  (t  exp[  ]
2 1 im 2 2ℏt 4( 1  im )
2iℏt (  )
2 2ℏt 2 2 2ℏt
2
 mx 
2  
 
1 m imx ℏt 
 exp[ ]
2 iℏt 
im
(2iℏt )
2 2 ℏt ℏ t  im  2
2( )
2ℏt  2 ℏt
2
 mx 
 
1 1 imx 2  ℏt 
 exp[  ]
2 iℏt
 2 2ℏt  
2 iℏt
m 2im( m)
 ℏt
2

2
 mx 
 
1 1 imx  ℏt  ℏt 
2 2
 exp[  ]
2 iℏt
 2 2 ℏt 2im  2 iℏt
m m

or

1 1 imx 2 imx 2  2
x  (t  exp(  )
2 iℏt
 2 2ℏt 2ℏt  2  iℏt
m m
1 1 imx 2 2
 exp[ (1  )]
2 iℏt
 2 2ℏt 2 
iℏt
m m
iℏt
1 1 imx 2
 exp[ ( m )]
2 iℏt
 2 2 ℏt 2 
iℏt
m m
2
1 1 x
 exp[ ]
2 iℏt  2

iℏt
 2 2 ( )
m m

Finally we get

1 1 x2
x  (t  exp[  ].
2 iℏt iℏt
2(  )
2
 
2

m m

97
It is clear that at t = 0, x  (t is the original Gaussian wave packet.

(b)

1 1 x2 iℏt
x  (t  exp[  ( 2  )]
2
2 2
iℏt ℏt m
2  2( 4  )
m m2
iℏt 2
x i
1 1  2 x2 m
 exp[  ] exp[ ]
2 iℏt ℏ2t 2 ℏ2t 2
 
2
2(  2 )
4
2(  2 )
4

m m m
ℏt
i x2
1 1  2 x2 m
 exp[  ]e i exp[ ]
2  4 ℏ t 
1/ 4 2 2
2 2 ℏ t ℏ2t 2
   2  2(  2 )
4
2(  2 )
4

m m m
 

where

iℏt ℏ2t 2 ℏ2t 2

2    4  2 e i with   arctan( ).
m m m 2 2

Then we have

2 1 1  2 x2
x  (t  exp[  ]
2 ℏ 2t 2 ℏ 2t 2
4  2 (  2 )
4

m m

2 1 1
The height of x  (t is .
2 ℏ 2t 2
4  2
m

ℏ 2t 2
1
The width is x    2 . 4

 m

22. Example-2: Feynman path integral

Suppose that the Gaussian wave packet is given by

1 x2
x  (t  0)  exp(ik 0 x  ). (Gaussian)
2  2 2

98
Here we discuss how such a Gaussian wave packet propagates along the x axis as the
time changes.

The free propagator:

m 1/ 2 im( x  x0 ) 2
K ( x, t; x0 , t0  0)  ( ) exp[ ].
2iℏt 2ℏt

Then we have

x  (t   dx 0 K ( x, t; x 0 ,0) x0  (0)
1/ 2
im( x  x 0 ) 2
2
1  m  x0
  
2   2iℏt   dx0 exp[ 2 2  2ℏt  ik0 x0 ]
Here we have

im( x  x0 ) 2 im( x 2  2 x 0 x  x 0 )
2 2 2
x x
 02   ik 0 x 0   0 2   ik 0 x 0
2 2ℏt 2 2ℏt
1 im mx imx 2
 ( 2  ) x0  i ( k 0  ) x0 
2

2 2ℏt ℏt 2ℏt
 ax0  bx 0  c
2

b 2 b 2  4ac
 a( x0  ) 
2a 4a

where

im1 mx imx 2
a 2  , b  i(k 0  ), c .
2 2ℏt ℏt 2ℏt

Then we get the integral

 
b 2 b2
 dx0 exp[ax0  bx0  c]  dx0 exp[a( x0  2a )  c  4a
2



b2 b
 exp(c  )  dx0 exp[a ( x0  ) 2
4a   2a
 b2
 exp( c  )
a 4a

99
 1
Note that when Re( a )    0 , the above integral can be calculated as
2 2

 
b 2 1 
 dx0 exp[a( x0  ) ]  dy exp(  y )
2
,

2a a 
a

b
with the replacement of variable as y   a ( x0  ) and dy  dx0  a . Thus we have
2a

1 m  b2
x  (t  exp( c  )
2  2iℏt  a 4a
2
 imx 
 ik 0  
1 m  imx 2
 ℏt 
 exp[  ]
2 2iℏt 1 im 2 2ℏt 4( 1  im )

2 2ℏt 2 2 2ℏt

or

2
 imx 
 ik 0  
1 m imx 2
 ℏt 
x  (t  exp[  ]
2 1 im 2 2ℏt 4( 1  im )
2iℏt (  )
2 2ℏt 2 2 2ℏt
2
 mx 
2   k0 
   ]
1 m imx ℏt
 exp[
2 iℏt 
im
(2iℏt 2 ) 2ℏt ℏt  im 2
2( )
2ℏt  2 ℏt
2
 mx 
2   k0 
   ]
1 1 imx ℏt
 exp[
2 iℏt
 2 2ℏt 2 
iℏt
m 2im( m)
 2 ℏt
2
 mx 
  k0 
1 1 imx 2
 ℏt  ℏt
2
 ]
 exp[ 
2 iℏt
 2 2 ℏt 2im  
2 iℏt
m m

or

100
 ℏk 0 t 2 2 2 iℏt
2 im( x  )  (  )
1 1 imx m m )
x  (t  exp( 
2
2 2
iℏt 2ℏt 2 ℏt ℏ t
 2 4  2
m m
1 1
 e  i
2  4 ℏ 2 t 2 
1/ 4

   2 
 m 
ℏk 0 t 2 ℏk 0 t 2
im( x  )  2 (x  )
imx2
m  4
1 m ]
exp(  2 2
) exp[ 
2 ℏt 2 ℏt ℏ t 2 ℏ 2t 2
4  4  2
m2 m

where

iℏt ℏ 2 t 2 i ℏ2t 2
 
2
   2 e
4
with   arctan( 2 2 ) .
m m m

Then we have

ℏk 0 t 2
) 2 (x 
2 1 1 m
x  (t  exp[  ].
2 ℏ 2 2
t ℏ 2 2
t
4  2 (  2 )
4

m m

This means the center of the Gaussian wave packet moves along the x axis at the constant
velocity.

2 1 1
(i) The height of x  (t is .
2 ℏ 2t 2
  2
4

1 ℏ 2t 2
(ii) The width is x  4  .
 m2

ℏk 0
(iii) The Group velocity is vg  .
m

23. Summary: Feynman path integral

The probability amplitude associated with the transition from the point ( xi , ti ) to
( x f , t f ) is the sum over all paths with the action as a phase angle, namely,

101
 i
Amplitude =  exp( ℏ S ) ,
All
paths

where S is the action associated with each path. So we can write down

K ( x f , t f ; xi , ti )  x f , t f xi , ti
tf
i
  exp(  dtL)
All ℏ ti
paths

i
 F (t f  ti ) exp( S cl )
ℏ

where Scl is the classical action associated with each path.

If the Lagrangian is given by the simple form

L( x, xɺ , t )  a(t ) xɺ 2  b(t ) xɺ x  c (t ) x 2 ,

then F (t f , ti ) can be expressed by

F (t f , ti )  K ( x f  0, t f ; xi  0, ti ) .

24. Comment by Roger Penrose on Feynman path integral

Roger Penrose: Road to the Reality. A Complete Guide to the Law of the Universe
(Jonathan Cape London, 2004).

Here is a very interesting comment by Roger Penrose (Nobel laureate, 2020) on the
Feynman Path integral. The content is the same, but some sentences are appropriately
revised.

The Lagrangian is in many respects more appropriate than a Hamiltonian when we

are concerned with a relativistic theory. The standard Schrödinger/Hamiltonian
quantization procedures lie uncomfortably with the spacetime symmetry of relativity.
However, unlike the Hamiltonian, which is associated with a choice of time coordinate,
the Lagrangian can be taken to be a completely relativistically invariant entity.
The basic idea, like so many of the ideas underlying the formalism of quantum
theory, is one that goes back to Dirac, although the person who carried it through as a
basis for relativistic quantum theory was Feynman. Accordingly, it is commonly referred
to as the formulation in terms of Feynman path integrals or Feynman sum over histories.

102
 The basic idea is a different perspective on the fundamental quantum mechanical
principle of complex linear superposition. Here, we think of that principle as applied, not
just to specific quantum states, but to entire spacetime histories. We tend to think of these
histories as ‘possible alternative classical trajectories’ (in configuration space). The idea
is that in the quantum world, instead of there being just one classical ‘reality’, represented
by one such trajectory (one history). There is a great complex superposition of all these
‘alternative realities’ (superposed alternative histories). Accordingly, each history is to be
assigned a complex weighting factor, which we refer to as an amplitude if the total is
normalized to modulus unity, so the squared modulus of an amplitude gives us a
probability. We are usually interested in amplitudes for getting from a point a to a point b
in configuration space.
The magic role of the Lagrangian is that it tells us what amplitude is to be assigned to
each such history. If we know the Lagrangian L, then we can obtain the action S, for that
history (the action being just the integral of L for that classical history). The complex
amplitude to be assigned to that particular history is then given by the deceptively simple
formula

i i
Complex amplitude  exp( S )  exp[  Ldt ] .
ℏ ℏ

For each history, there will be some action S, where S is the integral of the Lagrangian
along the path. All the histories are supposed to ‘coexist’ in quantum superposition, and
each history is assigned a complex amplitude exp(iS / ℏ) . How are we to make contact
with Lagrange’s requirement, perhaps just in some approximate sense, that there should
be a particular history singled out for which the action is indeed stationary?
The idea is that those histories within our superposition that are far away from a
‘stationary-action’ history will basically have their contributions cancel out with the
contributions from neighboring histories. This is because the changes in S that come
about when the history is varied will produce phase angles exp(iS / ℏ) that vary all
around the clock, and so will cancel out on the average. Only if the history is very close
to one for which the action is large and stationary (so the argument runs), will its
contribution begin to be reinforced by those of its neighbors, rather than cancelled by
them. because in this case there will be a large bunching of phase angles in the same
direction.
This is indeed a very beautiful idea. In accordance with the ‘path-integral’
philosophy, not only should we obtain the classical history as the major contributor to the
total amplitude—and therefore to the total probability—but also the smaller quantum
corrections to this classical behavior, arising from the histories that are not quite classical
and give contributions that do not quite cancel out, which may often be experimentally
observable.

103
25. Action in the Feynman path integral
The Feynman approach was inspired by Dirac’s paper (1933) on the role of the
Lagrangian and the least-action principle in quantum mechanics. This eventually led
Feynman to represent the propagator of the Schrödinger equation by the complex-valued
path integral which now bears his name. At the end of the 1940s Feynman (1950, 1951)
worked out, on the basis of the path integrals, a new formulation of quantum
electrodynamics and developed the well-known diagram technique for perturbation theory
We start with the Lagrange equation

d  Lcl  Lcl
  ,
dt  v  r

or

d L
P  cl ,
dt r

The conjugate momentum P can be derived as

Lcl 
P dt   Lcl dt ,
r r

or

 P  dr   L cl dt .

Here we use a type of techniques which is used in the Feynman-Hellmann theorem;

Hˆ 
 ( )  ( )   ( ) Hˆ  ( ) .
 

Thus, the action Scl can be expressed by

S cl   Lcl dt   P  dr .

in the Feynman path integral. The change in phase of the wave function is

104
 1 1
  Scl   P  dr .
ℏ ℏ

This simple form of the phase is essential point derived from the Feynman path integral.

26. Wave function for a charged particle in the presence of magnetic field
From the theory of the Feynman path integral, it is found that the change in phase of
the wave function is closely related to the action of classical Lagrangian, even for the
phenomena of quantum mechanics. The change in phase in the presence of a vector
potential A, is given by

1
ℏ
  P  dr ,

where P is the canonical momentum and p is the kinetic momentum, Lcl is the Lagrangian
and is defined by

1 q
Lcl  mv 2  A  v ,
2 c

where e is the charge of particle and m is the mass. The canonical momentum P is defined
by

Lcl q
P  mv  A .
v c

From the classical Lagrange theory. The equation of motion is governed by the Lagrange
equation,

d d L L
P  ( cl )  cl .
dt dt v r

The time derivative of the kinetic momentum p is equal to an external force F such as
Lorentz force,

dp
F, (Newton’s second law)
dt

105
which will be shown later. In quantum mechanics, it is known that the canonical
momentum P (but not p) is expressed by the differential operator

ℏ
P  .
i

We will discuss the application of the Feynman path integral in the next section 2S.
_______________________________________________________________________
REFERENCES
L.D. Landau and E.M. Lifshitz, The Classical Theory of Fields (Pergamon, 1996).
R.P. Feynman, Nobel Lecture December 11, 1965: The development of the space-time
view of quantum electrodynamics
http://www.nobelprize.org/nobel_prizes/physics/laureates/1965/feynman-
lecture.html
R.P. Feynman and A.R. Hibbs, Quantum Mechanics and Path Integrals (extended
edition) Dover 2005.
J.J. Sakurai and J. Napolitano, Modern Quantum Mechanics, second edition (Addison-
Wesley, New York, 2011).
A. Das; Lecture Notes on Quantum Mechanics, 2nd edition (Hindustan Book Agency,
2012).
John S. Townsend, A Modern Approach to Quantum Mechanics, second edition
(University Science Books, 2012).
R.P. Feynman, R.B. Leighton, and M. Sands, The Feynman Lectures on Physics vol. II
(Basic Book, 2010).
J. Mehra, The beat of different drum; The life and science of R. Feynman (Oxford
University Press, 1994).
H. Rauch and S.A. Werner, Neutron Interferrometry: Lessons in Experimental Quantum
Mechanics (Oxford, 2000).
K. Gottfried and T.-M. Yan, Quantum mechanics: Fundamentals, 2nd edition (Springer,
2003).
M. Reuter, Classical and Quantum Dynamics: From Classical Paths to Path Integral,
3rd edition (Springer, 2001).
L.S. Schulman, Techniques and Applications of Path Integration (Wiley-Interscience,
1981).
Colella, A. W. Overhauser, and S. A. Werner, Phys.. Rev. Lett. 34 ( 1975)
1472.”Observationof Gravitationally Induced Quantum Interference.”
E. Merzbacher, Quantum Mechanics, 3rd edition (John Wiley & Sons, 1998).
Roger Penrose: Road to the Reality. A Complete Guide to the Law of the Universe
(Jonathan Cape London, 2004).

_______________________________________________________________________
APPENDIX-I

Mathematical formula-1

106
 
 b2
 dx exp(ax  bx  c)   c)
2
exp(

a 4a

for Re[a]  0

((Mathematica))

________________________________________________________________________
APPENDIX-II Action in the classical mechanics
We start to discuss the calculus of variations with an action given by the form

tf

S   L[ xɺ, x ]dt ,
ti

dx
where xɺ  . The problem is to find has a stationary function xcl (x ) so as to minimize
dt
the value of the action S. The minimization process can be accomplished by introducing a
parameter .

107
 x

xf

x t

xcl t

xi

ti tf t

Fig.

x(ti )  xi , x (t f )  x f ,

x (t )  xcl (t )   (t ) ,

where  is a real number and

xcl (ti )  xi , xcl (t f )  x f ,

 (ti )  0 ,  (t f )  0 ,

 x 
x    d   (t )d ,
    0

108
 tf

S [ x ]   L( xcl (t )   (t ), xɺcl (t )  ɺ (t ))dt ,

ti

has a minimum at  = 0.

tf

S [ x  0 ]   L[ xcl (t )]dt ,
ti

 L[ x ]  L
   0, L  |  0 d ,
    0 

tf
S [ x ] L L
  {  (t )  ɺ (t )}dt
 ti
x xɺ
tf tf
L L d L
   (t )dt   (t )} |t if   ( ) (t )dt
t

ti
x xɺ ti
dt xɺ
tf
L d L
 [  ( )] (t )dt
ti
x dt xɺ
(1)

The Taylor expansion:

tf

S [ x ]   L( xcl (t )   (t ), xɺcl (t )  ɺ (t ))dt

ti

((Fundamental lemma))
If

tf

 M (t ) (t )dt  0
ti

for all arbitrary function (t) continuous through the second derivative, then M(t) must
identically vanish in the interval ti  t  t f .
________________________________________________________________________
From this fundamental lemma of variational and Eq.(1), we have Lagrange equation

109
 L d L
 ( )0. (2)
x dt xɺ

L can have a stationary value only if the Lagrange equation is valid.

In summary,

x2

S   L( x, xɺ )dt ,
x1

L d L
S = 0   ( )0.
x dt xɺ

APPENDIX-III: Lagrangian of charged particle in the electromagnetic field

(III-1) Lagrangian of free particle

((L.D. Landau and E.M. Lifshitz))

In the course of an infinitesimal time interval dt the moving clocks go a distance dr. Let us
ask what time interval (proper time d ) is indicated for this period by the moving clocks. In a
system of coordinates linked to the moving clocks, the latter are at rest, dr '  0 . Because of the
invariance of intervals

c 2 (d ) 2  (dr ') 2  c 2 (d )2  c 2 (dt ) 2  (dr )2

or

1
d  ( dt ) 2  ( dr ) 2
c2
1 dr 2
 dt 1  ( )
c 2 dt
v2
 dt 1 
c2

dr
where v  is the velocity. Integrating this expression, we can obtain the time interval
dt
indicated by the moving clocks when the elapsed time according to a clock at rest is  . The
action S is given by

110
 v2
S   Ldt  mc 2  d  mc 2  dt 1  ,
c2

yielding to the expression of the Lagrangian for the free particle as

v2
L  mc 2
1 2 .
c

where m is the mass of the particle. Note that the units of L is erg.

(III-2) Four-potential of the electromagnetic field.

The four-dimensional vector for the scalar potential and vector potential,

A  ( , A ) , (Gaussian)


A   ( , A) . (SI units)
c

Hereafter, we use the Gaussian units for the four-dimension potential. We note that

x   (ct , r ) , x  (ct , r ) ,

and

x  x  c 2 t 2  r 2 .

Generalized potential U is obtained as

q
Udt  A dx 
c
q
 (c dt  A  dr )
c
1 dr
 q ( dt  A  dt )
c dt
1
 q (  A  v) dt
c

or

111
 1
U  q (  A  v ) .
c

where q is a charge of particle. The Lagrangian for a charged particle in an electromagnetic field
has the form

L  K U
v2 1
 mc 2
1  2  q(  A  v )
c c

The action S is expressed by

t t v2 q
S   Ldt1   (mc 2 1   A  v  q )dt1
t' t' c2 c

The generalized momentum is

L mv q q
P   A  p A,
v v2 c c
1
c2
with

mv
p .
v2
1 2
c

We also note that

S cl   P  dr . (Feynman path integral)

The Hamiltonian H is obtained as follows.

112
 H  Pv L
q v2 1
 (p  A)  v  mc 1  2  q(  A  v )
2

c c c
mv 2 v2
  mc 2 1   q
v2 c2
1
c2
mc 2
  q
v2
1 2
c

or

H  q 2 m 2 c 2
( ) 
c v2
1 2
c
v2
m 2 c 2 (1 
2
)  m 2v 2
 c
v2
1 2
c
 m c c p
2 2 2 2

q 2
 m 2c 2  (P  A)
c

or else

q
H  m 2 c 4  c 2 (P  A )2  q
c

For low velocities, i.e., for classical mechanics, the Lagrangian goes over into

1 2 q
L mv  A  v  q
2 c

and the Hamiltonian is

113
 H  Pv L
mv 2 v2
  mc 2
1  2  q
v2 c
1 2
c
2
mc
  q
v2
1 2
c

In this approximation,

q
p  mv  P  A
c

The final form of the Hamiltonian is given by

1 2 1 q
Hɶ  H  mc 2  p  q  ( P  A) 2  q
2m 2m c

(III-3) Derivation of Lorentz force from Lagrangian

The Lorentz force, which is a force on a particle with a charge q due to an electric field E and
magnetic field B,

1
F  q ( E  v  B)
c

where v is the velocity of particle

1 A
E    , (electric field)
c t

B   A . (magnetic field)

The gauge transformation is defined by

1 
A '  A   , ' 
c t

where  is an arbitrary function. Suppose that the Lagrangian L is expressed as

114
 v2 1
L  mc 2 1  2
 q(  A  v )
c c
1 1
 mv 2  q(  A  v )
2 c
 T U

is the velocity-dependent potential energy. Note that

L mv q q
P   Ap A
v v2 c c
1
c2

Lagrange equation is

d  L  L
   L ,
dt  v  r

with

q
L   (  q  v  A) ,
c

d
where is a total derivative. We note that
dt

( v  A)  ( v ) A  ( A ) v
 v  (  A)  A  (  v )

Since v and r are independent variables in the Lagrangian, we have

( A ) v  0 , A  (  v )  0

leading to

( v  A)  ( v  ) A  v  (  A)

We note that

115
 d  L  d q
   (p  A )
dt  v  dt c
dp q dA
 
dt c dt
dp q A
  [  ( v )A ]
dt c t

with

dAi Ai A A A
  xɺ i  yɺ i  zɺ i
dt t x y z
A
 i  ( v  ) Ai
t

or

dA A
  ( v  ) A]
dt t

So that, we have

q
L   q   ( v  A)
c

We get the final form of the Lagrange equation as

dp q A q
 [  ( v ) A ]   q  ( v  A )
dt c t c

or

dp 1 A q
 q (  )  [( v  A)  ( v ) A]
dt c t c
1 A q
 q (  )  v  (  A)
c t c
1
 q (E  v  B )
c

which is equivalent to the Lorentz force, with

116
 ( v  A)  ( v ) A  v  (  A)  v  B

(III-4) Gauge transformation

The original Lagrangian L0 is changed into a new Lagrangian L0 ' during the Gauge
transformation,

1
L0 '   q( ' A ' v)
c
1  q
  q(  )  ( A    v )
c t c
q 
 L0  (    v)
c t
q d  (r , t )
 L0 
c dt

But as we know, adding to the Lagrangian a total time derivative of a function of r and t does not
d
change the equations of motion. The function always satisfies the Lagrange equation. Note
dt
that

d  d d   d
[ ] [ ]
dt rɺ dt dt r r dt

since

d     
  xɺ  yɺ  zɺ
dt t x y z

  ( v ) 
t

d  d d   d
[ ] [ ] ,
dt xɺ dt dt x x dt

d  d d   d
[ ] [ ] ,
dt yɺ dt dt y y dt

d  d d   d
[ ] [ ] .
dt zɺ dt dt z z dt

117
In other words,

d  
( L ')  L'
dt rɺ r

is equal to

d  
( L0 )  L0
dt rɺ r

d
independent of the form .
dt

((Comment))
Yasushi Takahashi (Note on Mathematical Physics, Kodansha, 1992) in Japanese.

"It may be reasonable to use the expression of the Lagrangian (classical system) given by

1 2 q
L mv  A  v  q .
2 c

The reason is why the Lagrange equation for this form of Lagrangian yields the equation of
motion of the charged particles with charge q with the Lorentz force F,

dp 1
 F  q (E  v  B ) .
dt c

(III-5) Feynman path integral

q
S cl   P  dr   (p  A )  dr
c

For the closed path

 A  dr   (  A)  da
  B  da
 B

118
((Stokes’ theorem))

1 q q
ℏ  c A  d r  cℏ  B

Magnetic quantum flux (fluxoid)

hc 2 ℏc  ℏc
0    .
2e 2e e

 0  2.0678  107 Gauss cm2 (cgs units)

or

 0  2.0678  1015 T m2 (SI units)

REFERENCES
R. Shankar, Principles of Quantum Mechanics, 2nd edition (Kluwer Academic/Plenum
Publisher, 1994).
H. Goldstein, C.P. Poole, and J. Safko, Classical Mechanics, 3rd edition (Pearson, 2011).
L.D. Landau and E.M. Lifshitz, The Classical Theory of Fields, 4th revised English edition,
Course of Theoretical Physics volume 2 (Pergamon, 1975).
C. Kittel, Introduction to Solid State Physics, 8th edition (John Wiley and Sons, 2005).
Yasushi Takahashi, Note on Mathematical Physics, Kodansha, 1992) in Japanese.

119