The Classical Limit of Quantum Mechanics

If the theory of quantum mechanics is correct, then when quantum mechanics is applied to big things, it
must give the results of classical mechanics.

Because quantum mechanics emphasizes potential energy V (x), and classical mechanics emphasizes force
F , let’s remind ourselves how they’re related. The definition of potential energy (in one dimension) is
Z x
V (x) − V (x0 ) = − F (x0 ) dx0 , (1)
x0

where F (x) is the force function — i.e., the force that would be exerted on the particle if the particle were at
position x. (I’m not saying that the particle is at x, I’m saying that’s what the force would be if the particle
were at x.) Taking the derivative of both sides with respect to x (and using the fundamental theorem of
calculus on the right)
dV (x)
= −F (x). (2)
dx
In the classical limit, the quantal spread in x will be small. We ask how the expected position (the
“center of the wavepacket”) moves:
Z +∞
hxi = xψ ∗ (x, t)ψ(x, t) dx, (3)
−∞

so +∞ +∞
∂ψ ∗ (x, t)
Z Z
dhxi ∂ψ(x, t)
= x ψ(x, t) dx + xψ ∗ (x, t) dx. (4)
dt −∞ ∂t −∞ ∂t
But the Schrödinger equation tells us how wavefunction ψ(x, t) changes with time:

h̄2 ∂ 2 ψ(x, t)
 
∂ψ(x, t) i
=− − + V (x)ψ(x, t) (5)
∂t h̄ 2m ∂x2
and
∂ψ ∗ (x, t) h̄2 ∂ 2 ψ ∗ (x, t)
 
i ∗
=+ − + V (x)ψ (x, t) . (6)
∂t h̄ 2m ∂x2
(From here on I’m going to write ψ(x, t) as ψ and V (x) as V .) Thus

h̄2 ∂ 2 ψ ∗
Z +∞  
dhxi i ∗
= − x − + V ψ ψ dx
dt h̄ −∞ 2m ∂x2
h̄2 ∂ 2 ψ
Z +∞   
∗
− xψ − + V ψ dx (7)
−∞ 2m ∂x2
h̄2 +∞ Z +∞
∂ 2 ψ∗ 2
 Z 
i ∗∂ ψ
= − − x ψ dx − xψ dx
h̄ 2m −∞ ∂x2 −∞ ∂x2
Z +∞ Z +∞ 
+ xV ψ ∗ ψ dx − xψ ∗ V ψ dx (8)
−∞ −∞
Z +∞ 2 ∗ Z +∞ 2

h̄ ∂ ψ ∗∂ ψ
= i x ψ dx − xψ dx (9)
2m −∞ ∂x2 −∞ ∂x2

1
 Can we say anything about integrals such as the second integral in square brackets above? Surprisingly,
the answer is yes. If we define
∂ψ
f (x) = xψ ∗ and g(x) = (10)
∂x
then Z +∞ Z +∞
∂2ψ
xψ ∗ 2 dx = f (x)g 0 (x) dx (11)
−∞ ∂x −∞
which suggests integration by parts:
Z +∞ Z +∞
0 +∞
f (x)g (x) dx = [f (x)g(x)]−∞ − f 0 (x)g(x) dx. (12)
−∞ −∞

Now remember what the wavefunction is normalized, so it has to fall to zero at both infinity and negative
infinity. Typically the slope ∂ψ/∂x also falls to zero at both infinity and negative infinity, and does so very
rapidly — much more rapidly than linearly. (There are exceptions to these typical behaviors, such as the
Coulomb scattering wave functions, and in these atypical cases this argument has to be rethought.) The
upshot is that in typical situations
+∞
[f (x)g(x)]−∞ = 0 (13)

so +∞ +∞
∂2ψ ∂(xψ ∗ ) ∂ψ
Z Z
xψ ∗ dx = − dx. (14)
−∞ ∂x2 −∞ ∂x ∂x
We’ll use this trick several times. . . I’ll just call it the “integration-by-parts trick”.

Applying this trick to both integrals of equation (9) gives

Z +∞ Z +∞
∂(xψ) ∂ψ ∗ ∂(xψ ∗ ) ∂ψ

dhxi h̄
= −i dx − dx
dt 2m −∞ ∂x ∂x −∞ ∂x ∂x
Z +∞ Z +∞ Z +∞ Z +∞
∂ψ ∂ψ ∗ ∂ψ ∗ ∂ψ ∗ ∂ψ

h̄ ∂ψ
= −i x dx + ψ dx − x dx − ψ∗ dx
2m −∞ ∂x ∂x −∞ ∂x −∞ ∂x ∂x −∞ ∂x
 Z +∞ Z +∞
∂ψ ∗

h̄ ∂ψ
= −i + ψ dx − ψ∗ dx
2m −∞ ∂x −∞ ∂x
Z +∞ 
h̄ ∂ψ
= − =m ψ∗ dx . (15)
m −∞ ∂x

Notice that dhxi/dt is real, as it must be. (This isn’t proof that we’ve made no algebra errors, but if our
expression for dhxi/dt had been complex, that would have been proof that we had made algebra errors.)

All this is fine and good, but it takes us only part way to our goal. This is clearly not a classical
equation. . . it contains h̄ right there! Since the classical F = ma involves the second derivative of position
with respect to time, we take one more derivative of hxi.

+∞ +∞
d2 hxi ∂ψ ∗ ∂ψ
Z Z 
h̄ ∂ ∂ψ
= − =m dx + ψ∗ dx . (16)
dt2 m −∞ ∂t ∂x −∞ ∂x ∂t

2
The second-order derivative on the right looks particularly grotesque, so we use the integration-by-parts
trick to get rid of it:
Z +∞ Z +∞
d2 hxi ∂ψ ∗ ∂ψ ∂ψ ∗ ∂ψ

h̄
= − =m dx − dx (17)
dt2 m −∞ ∂t ∂x −∞ ∂x ∂t
Z +∞
∂ψ ∗ ∂ψ

2h̄
= =m dx . (18)
m −∞ ∂x ∂t
Now use Schrödinger’s equation:
d2 hxi +∞
∂ψ ∗ h̄2 ∂ 2 ψ
Z    
2h̄ i
= =m − − + V ψ dx
dt2 m −∞ ∂x h̄ 2m ∂x2
+∞
∂ψ ∗ h̄2 ∂ 2 ψ
Z   
2
= − <e − + V ψ dx . (19)
m −∞ ∂x 2m ∂x2
Look at that. . . two of the h̄s have cancelled out! We’re not home yet because there’s still an h̄ within the
square brackets, but we’re certainly making progress. We have that
h̄2
Z +∞ Z +∞
d2 hxi ∂ψ ∗ ∂ 2 ψ ∂ψ ∗
 
2
= − <e − dx + V ψ dx , (20)
dt2 m 2m −∞ ∂x ∂x2 −∞ ∂x
but let’s apply the integration-by-parts trick to the first integral:
Z +∞ Z +∞ 2 ∗
∂ψ ∗ ∂ 2 ψ ∂ ψ ∂ψ
2
dx = − dx. (21)
−∞ ∂x ∂x −∞ ∂x2 ∂x
Think about this for a minute: if the integral on the left is z, this equation says that z = −z ∗ , whence z is
pure imaginary or <e{z} = 0. Thus
Z +∞
d2 hxi ∂ψ ∗

2
= − <e V ψ dx , (22)
dt2 m −∞ ∂x
an expression devoid of h̄s! Apply the integration-by-parts trick to this integral:
Z +∞ Z +∞
∂ψ ∗ ∂(V ψ)
V ψ dx = − ψ∗ dx
−∞ ∂x −∞ ∂x
Z +∞ Z +∞ Z +∞
∂ψ ∗ ∂ψ ∂V
V ψ dx = − ψ∗ V dx − ψ∗ ψ dx
−∞ ∂x −∞ ∂x −∞ ∂x
Z +∞ Z +∞ Z +∞
∂ψ ∗ ∗ ∂ψ ∂V
V ψ dx + ψ V dx = − ψ∗ ψ dx
−∞ ∂x −∞ ∂x −∞ ∂x
Z +∞ +∞
∂ψ ∗
 Z
∂V
2<e V ψ dx = − ψ∗ ψ dx (23)
−∞ ∂x −∞ ∂x
Plugging this result back into equation (22) gives
+∞
d2 hxi
Z
1 ∂V
= − ψ∗ ψ dx. (24)
dt2 m −∞ ∂x
But remember that the force function is F (x) = −∂V /∂x, so
d2 hxi 1 +∞
Z
1
2
= F (x)ψ ∗ (x, t)ψ(x, t) dx = hF (x)i. (25)
dt m −∞ m

3
There it is —
d2 hxi
hF (x)i = m . (26)
dt2
This result is called Ehrenfest’s theorem.

Note that the result is not

d2 hxi
F (hxi) = m . (27)
dt2