Ehrenfest’s Theorem

Ehrenfest’s theorem simply states that expectation values of quantum mechanical

operators obey the laws of classical mechanics. Classically, the hamiltonian for a
single particle of mass m,
p~2
H = + V (~r)
2m
leads to the equations of motion,
dx ∂H px dpx ∂H ∂V
= = , = − = − . (24)
dt ∂px m dt ∂x ∂x
In quantum mechanics we can calculate the rate of change of expectation value of x,
Z
dhxi ∂
= d~r ψ ? (~r, t) x ψ(~r, t)
dt ∂t
Z  ? 
∂ψ ? ∂ψ
= d~r xψ + ψ x
∂t ∂t
Z
i~
d~r − ∇2 ψ ? x ψ + ψ ? x ∇2 ψ .



= (25)
2m
In the last step (25) we have made use of Schrödinger equation for ψ ? and ψ. Upon
partial integration, we observe that,
Z Z Z
2 ?
dτ (∇ ψ )xψ = dτ ∇ · (∇ψ xψ) − dτ ∇ψ ? ∇(xψ)
?

Z Z Z
= dτ ∇ · (∇ψ xψ) − dτ ∇ · (ψ ∇( xψ)) + dτ ψ ? ∇2 (xψ).
? ?

Using divergence theorem, the first two integrals can be converted to surface integrals
and can be made to vanish on suitably choosen large surfaces using the well-behaved
nature of the wave functions. As for the third term, we notice that,
∇2 (x ψ) = 2x̂ · ∇ψ + x ∇2 ψ
where x̂ in this context is the unit vector in x-direction. Using this expression in the
equation (25), we get,
Z
dhxi i~
d~r −2ψ ? x̂ · ∇ψ − ψ ? x ∇2 ψ + ψ ? x ∇2 ψ
 
=
dt 2m
Z
1
= d~r ψ ? x̂ · (−i~ ∇) ψ
m
Z  
1 ? ∂
= d~r ψ −i~ ψ
m ∂x
Z
1
= d~r ψ ? p̂x ψ.
m
Therefore, instead of classical expression, we have for the first equation of motion,
dhxi hpx i
= . (26)
dt m
1
In the same way, using Schrödinger equation and divergence theorem, we can show
 
dhpx i ∂V
= − . (27)
dt ∂x
In general terms, the time variation of expectation value of any arbitrary quantum
mechanical operator (or dynamical variable) is given by,
Z
dhαi d
= d~r ψ ? α̂ ψ
dt dt
Z  ? 
∂ψ ? ∂α ? ∂ψ
= d~r αψ +ψ ψ+ψ α
∂t ∂t ∂t
?
  Z     
∂α 1 ∂ψ ? ∂ψ
= + d~r − −i~ α ψ + ψ α i~ (28)
∂t i~ ∂t ∂t
Making use of time-dependent Schrödinger equation (5), Ĥ ψ = i~∂ψ/∂t and its
complex counterpart, we arrive at
  Z
dhαi ∂α 1
= + d~r ψ ? (αĤ − Ĥα)ψ
dt ∂t i~
 
dhαi ∂α 1 D E
= + [α, Ĥ] . (29)
dt ∂t i~
The above equation (29) is known as Heisenberg’s equation of motion and the quantity
[α, Ĥ] is called commutator of α and Ĥ. If two different observables or dynamical
variables be represented by the operators  and B̂, then the commutator of  and
B̂ is defined as
ÂB̂ − B̂ Â = [Â, B̂]. (30)
Heisenberg’s equation of motion (29) closely resembles the corresponding classical
equation of motion in terms of Poisson bracket,
!
∂ Â ∂ B̂ ∂ Â ∂ B̂
[Â, B̂]pb = − (31)
∂q ∂p ∂p ∂q
which is,
dA ∂A
= + [A, H]pb . (32)
dt ∂t

Hamiltonian mechanics in nut-shell

The equation of motion of a system described by Lagrangian L(q, q̇, t) is given by Euler-
Lagrange equation and the canonical conjugate momenta pi for each generalized coordinate
qi are defined as,  
d ∂L ∂L ∂L
− = 0, pi = . (33)
dt ∂ q̇ ∂q ∂ q̇i
Equivalently, a physical system can be described by generalized cooredinates and momenta
in terms of Hamiltonian H(q, p), and correspondingly we get Hamiltonian equations,
X ∂H ∂H
H(q, p) = q̇i pi − L (Legendre transformation), q̇i = , ṗi = − . (34)
∂pi ∂qi
i

2
In terms of Poisson bracket, q̇, ṗ and any arbitrary dynamical variable f , the equation of
motions are,
df ∂f
q̇i = [qi , H]pb , ṗi = [pi , H]pb and = + [f, H]pb . (35)
dt ∂t
However, the choices of generalized coordinates and momenta are not unique and one choice
{Qi , Pi } may be prefered over the other {qi , pi }. For instance, to solve Kepler’s problem,
(r, θ) coordinates are prefered over (x, y) because of cyclic nature of θ in the problem,
which actually helps in identifying conserved quantity(s) rather easily. Transformation of
generalized coordinates (q, p) → (Q, P ) that will preserve the form of Hamilton’s equation,
while the explicit form of Hamiltonian differ, is known as the canonical transformation,
∂K ∂K
H(q, p) → K(Q, P ) : Q̇i = , Ṗi = − . (36)
∂Pi ∂Qi
Since Hamilton’s principe must be obeyed,
Z t2
δ [q̇p − H(q, p)] dt = 0 (37)
t1
Z t2 h i
δ Q̇P − K(Q, P ) dt = 0. (38)
t1

Since the general form of Hamilton’s principle has zero variation at the end points, a con-
sistent way to connect K with H is through generating function F ,
dF
q̇p − H = Q̇P − K + , (39)
dt
where F is the generating function of one of the old canonical coordinates (q, p) and one
new (Q, P ). Therefore, we can have 4 types of generating functions (making use of Legendre
transformation),

type 1 : F = F1 (q, Q), (40)

type 2 : F = F2 (q, P ) − QP, (41)
type 3 : F = F3 (p, Q) + qP, (42)
type 4 : F = F4 (p, P ) + qP − QP. (43)

Consider the type-2 generating function and insert in (39),

dF2
q̇p − H = Q̇P − K − Q̇P − QṖ +
dt
∂F2 ∂F2 ∂F2
= −QṖ − K + q̇ + Ṗ + . (44)
∂q ∂P ∂t
Comparing lhs with rhs, we obtain,
∂F2 ∂F2 ∂F2
p = , Q = and K = H+ . (45)
∂q ∂P ∂t
Identifying conserved quantities and constants of motion of a mechanical system can be
trivial when,
 
∂S ∂S
F2 (q, P ) ≡ S(q, P ) ⇒ K = H q, + = 0. (46)
∂q ∂t

3
The above equation is known as Hamilton-Jacobi equation and S is also known as Hamilton’s
principal function. Therefore, Q and P , the new generalized coordinate and momenta are
constants of motion that follows from (36),

∂K ∂K
Q̇ = = 0 and Ṗ = − = 0. (47)
∂P ∂Q
In fact, Hamilton-Jacobi equation yields all constants of motion automatically and the
solutions can be expressed in terms of these constants. Another interesting feature of this
equation is that Hamilton’s principal function S behaves like a wave (or wave-front), since it
can be shown that particle trajectories pierce surfaces of constant S. Hence the Hamilton-
Jacobi is considered to be the only formulation of mechanics in which the motion of a
particle can be represented as a wave.

Classical limit
From Ehrenfest’s theorem we came to know that the expectation values of the
quantum mechanical observables follow laws of classical mechanics. Here, we discuss
the classical limit, ~ → 0 of Schrödinger equation. Let the wave function be written
as, p
ψ(~r, t) = P (~r, t) eiS(~r,t)/~ ⇒ ψ ? (~r, t) ψ(~r, t) = P (~r, t) (48)
which is consistent with equation (12). The physical interpretation of S(~r, t) follows
from (13) by noting,
√ √ i √ √ i 2i
ψ ? ∇ ψ − ∇ψ ? ψ = P ∇ P + P ∇S − P ∇ P + P ∇S = P ∇S.
~ ~ ~
Therefore, the quantity S is related to the current density,

~j = P ∇S ⇒ p~ = ∇S → ~v = ∇S . (49)
m m
√
Substituting ψ = P exp(iS/~) in time-dependent Schrödinger equation (5),
" √ #
∂ P i √ ∂S
i~ + P = (50)
∂t ~ ∂t
~2 √ 2i √ 1√ i√ √
 
2 2 2
− ∇ P + ∇ P · ∇S − 2 P (∇S) + P ∇ S + P V (51)
2m ~ ~ ~

Here we regard ~ to be a small quantity and can drop terms in (51) that contain ~
to obtain,

1 ∂S(~r, t) p2 ∂S(~r, t) ∂S
[∇S(~r, t)]2 + V (~r) + = +V + = H+ = 0. (52)
2m ∂t 2m ∂t ∂t
The equation (52) is just the Hamilton-Jacobi equation in classical mechanics, the S
being type-2 generating function (S(q, P, t)).