8.

04 Quantum Physics

Lecture XIII

p = p h = dx i x h = dx i x h = dx ( ) i x x h = dx i x = p ,

(13-1) (13-2) (13-3) (13-4) (13-5) (13-6)

where again we have used integration by parts and the fact that vanishes at . Consequently, p = p , i.e. all expectation values of p are real. Since an eigenvalue is the expectation value for the corresponding stationary state, all eigenvalues of the momentum operator must be real. An operator whose eigenvaluse are real (or equivalently, whose expectation value for all admissible wavefunctions is real) is called a Hermitian operator. Physically measurable quantities are represented by Hermitian operators. Similarly, one can show that E = E for any state, so all energy eigenvalue are is a Hermitian operator. real: The Hamiltonian operator H

Can we derive Newtons F = ma from the SE?

CM: F = dV = ma = dx dp dt =
dp dt

Let us calculate the expectation value of

dp : dt

d p dt d h = dx (x, t) (x, t) dt i x h = dx + i t x x t h 2 2 h 2 2 = dx +V +V 2m x2 x x x 2m x2
A

(13-7) (13-8) (13-9) (13-10)

Massachusetts Institute of Technology

XIII-1

8.04 Quantum Physics The integrand is 2 h 2 2 h 2 A= +V + 2m x x2 2m x2 x x h 2 2 V V 2 x x 2m x x 2 h V = 2m x x x x x

Lecture XIII

(13-11) (13-12)

Again the integral over the rst term vanishes since 0 for x , and we are left with

dp dt

V dV dx (x, t) (x) (x, t) = x dx

(13-13)

or d p = dt dV dx (13-14)

It follows from the SE that the expectation values obey the classical equations of motion. m d x = p dt d dV p = dt dx (13-15) (13-16)

Average momentum changes due to average force dV V = dx = dxF (x)|(x, t)|2 , x dx

(13-17)

i.e. position-dependent force F (x) is weighted by probability density x, t)|2 for dV |( d nding the particle at position x at time t. Note, however, that dx = dx V (x) Example 1. Double-peaked distribution. The probability to nd the particle at the average position x is small, so the force there cannot be of much consequence for the particles motion. Example 2. Force varying quickly on wavepacket scale. Classical calculation ddx V (x) would predict very large (and quickly varying force as x changes), actual QM force Massachusetts Institute of Technology XIII-2

8.04 Quantum Physics

Lecture XIII

Figure I: Double-peaked particle distribution with vanishing probability to nd particle at average position.

Figure II: Particle wavepacket large compared to spatial variation of the force. The time evolution of the wavepacket will depend on the value of the force averaged over the wavepacket, not just on the force at the average particle location. dV experienced by particle is much smaller. However, if the force varies slowly dx compared to the size of a single wavepacket, then dV d = F (x) F (x) = V (x) (13-18) d x dx This is the reason why we can treat particles in macroscopic potential usually as classical particles. d dV p = always true (13-19) dt dx Massachusetts Institute of Technology XIII-3

8.04 Quantum Physics d d p V (x) dt d x

Lecture XIII Ehrenfests theorem for slowly varying potentials (13-20)

Eigenfunctions of the momentum operator

What are the eigenfunctions up of the momentum operator, i.e. the eigenfunctions satisfying pu p = pup , (13-21)

where p is some (xed) particular eigenvalue of p . We know that the operator p is h Hermitian, so all eigenvalues p are real. In position space, we have p = i x and h ipx/h u ( x ) = pu ( x ), u ( x ) = Ae . The momentum eigenfunctions are (of course) p p i x p just the plane waves. Let us check the orthonormality condition for eigenstates:

dxup (x)up (x) = A

Ap

ipx/h dxeip x/h e h dxei(pp )x/

(13-22) (13-23) (13-24) (13-25) (13-26)

= Ap Ap

=h Ap Ap

dyei(pp )y

=h Ap Ap (p p )2 =h 2Ap Ap (p p )

The momentum eigenfunction are orthogonal for p = p , but we have a normalization problem for p = p : The Dirac delta function diverfes, or equivalently, the integral dx|up (x)| = |AP |
2 2

h 2 dx |eipx/ | =1

(13-27)

diverges. Before looking at possibilities to deal with normalization problem, let us calculate the expansion coecients c(p) c(p) = dxAp e
ipx/ h

(x) = Ap

h dx (x)eipx 2 h(p)

(13-28)

Massachusetts Institute of Technology

XIII-4

8.04 Quantum Physics

Lecture XIII

1 We see that if we make the normalization choice Ap = 2 , then the expansion h coecients c(p) into momentum eigenstates are just given by the Fourier transform

up (x) = (x) =

1 h eipx/ 2h

normalized momentum eigenstates

(13-29) (13-30)

expansion into momentum eigenstates

dp(p)up (x)

1 = 2h

Fourier transformation

h dp(p)eipx/

(13-31)

The expansion into momentum eigenstates and Fourier tranformation are one and the same. Since (x) and (p) contain the same information about the particle, we can use either one to characterize the position and motion of the particle. A more fundamental motion is the state of the particle (a state is a vector in Hilbert space), the state can be expressed (written down) in various representations (like position representation (x), momentum representation (p), energy representation ). We cE ) associated with Hermitian operators (position x , momentum p , energy H call (p) the momentum representation of a particular state, and interpret it as the wavefunction in momentum space. The SE governs the time evolution of the wavefunction, or equivalently, the time evolution of the state of the particle in Hilbert space. For one particle in one (three) dimensions, the Hilbert space is one- (three-) dimensional, but for N particles in three dimensions the Hilbert space is 3N -dimensional. In general, it cannot be factored into a tensor product of N three-dimensional vector space Vsystem = V1 V2 VN , or equivalently, the wavefunction for N particles does not factor into a product of wavefunctions for each particle, 1 (r1 )2 (r2 ) . . . N (rN ) (r1 , r2 , . . . , r1 , t) = (13-32)

In this case, when the wavefunction for an N -particle system cannot be written as a product of wavefunctions for the individual particles, i.e. when the particles do not evolve independently, we speak of an entangled state. Because of this possibility a quantum system of N particles is vastly (exponentially in N ) richer than an classical system of N particles. However, in most cases we lose track of the particle-particle correlations associated with entanglement, and the system behaves quasi-classically. A quantum system that could preserve the correlations, and that could be manipulated externally, would constitute a quantum computer. A quantum computer could solve certain computation problems (only a handful have been discovered so far) exponentially faster than a classical computer. Because of the enormous size of the Massachusetts Institute of Technology XIII-5

8.04 Quantum Physics

Lecture XIII

Hilbert space, certain quantum mechanical problems involving many-particle correlations (e.g. high temperature superconductivity that involves correlated motion of many electrons) are very dicult to solve or simulate on a classical computer. Now back to a single particle in one dimension . . .

Massachusetts Institute of Technology

XIII-6