Anda di halaman 1dari 8

Ph125b Wednesday 3 January 2007

In this part of the course well move on from finite-dimensional quantum systems to the motion of a point particle in one kinetic dimension. What we mean by one dimension here is that the particle is constrained to move along a line, as opposed to being free to move throughout three-space. Classically we would describe its position by a single variable x, as opposed to a position vector x, y, z. r As we are about to see, in quantum mechanics the state of a point particle even in one kinetic dimension corresponds to a vector in an infinite-dimensional complex Hilbert space.

Quantum Mechanics in Infinite Dimensions


Formally, we may represent the motional state of a point particle (in one kinetic dimension) by a ket in an infinite-dimensional Hilbert space, | H. One way to understand the need for an infinite-dimensional Hilbert space is to think about what should constitute a complete measurement. If this is truly a point particle, then every possible value of position x should correspond to an orthogonal projection operator, or equivalently to one member of a complete orthogonal basis set for the Hilbert space. Since x is a continuous variable, it would seem that the Hilbert space should be infinite dimensional. Using | x to denote the basis ket corresponding to a particular value of position, we find that it should take an infinite number of complex coefficients to fully specify an arbitrary ket | H : x | x, where the continuous function x here is known as the wave function corresponding to the state | . Just as we often used vector representations for bras and kets in finite dimensions, we will almost always use x to perform computations in infinite dimensions. Note that | x x | x, and here in infinite dimensions we must replace the usual type of closure relation

| j j | 1
j1

with an integral

dx | x x | 1.
Hence

1 | 2 1 | and the norm of a state is |


dx | x x |

| 2

dx 1 | x x | 2 dx x 2 x 1 dx | x | 2 .

We would like to require this norm to be equal to one for all valid quantum states, but below well see that the issue of what constitutes a valid wave-function can get subtle. The notion of orthogonality is now 1 | 2

dx x 2 x 0, 1 cdj 0 j
j

which we can compare with finite-dimensional expressions like

where | 1

c j | j ,
j

| 2

d j | j .
j

Note also that if we really want to have x | x, we are forced to conclude that the wave function corresponding to a position basis ket | x is the singular object | x x x , so that x | | x

dx x x x x , dx x x x x . dx x x x x 1,

Its not exactly clear what we would mean by the requirement x | x

so we already have our first instance of a wave function wed like to make use of (in one way or another!) that is not square-normalizable. The outer product | | of a state with itself is still an operator, with matrix elements x || || x x | | x x x. The most important new subtlety that arises in the interpretation of | is that | x | | x | | 2 | x | 2 must be interpreted not as the probability to find the particle precisely at position x, but rather as a probability density in the vicinity of position x. Hence, we should think of

dP | x | 2 dx as an infinitesimal probability to find the particle within dx of position x, and Prx 1 x x 2

x2
1

dx | x | 2

as the finite probability to find the particle within the interval x 1 x x 2 . Now we are finally ready to introduce the position observable, x. We would like this operator to have eigenkets corresponding to the position basis states introduced above, x | x x | x . Hence the expectation value of position with respect to a given state | must be |x|

dx | x x | x dx x x x,

where in the first line we inserted a spectral decomposition (note that for the moment we are sweeping some technical subtleties under the rug), x Likewise | xn |

dx | x x | x. dx x x n x,

where our reason for writing x n between x and x will soon be made clear.

From classical mechanics we know that a complete kinetic description of particle motion must include momentum as well as position. Thus we have a momentum operator p, which we define by its matrix elements 1 | p | 2 dx x i 2 x 1 x 2 x . i dx x 1 x Hence we make the association p i x when working with wave-functions in one dimension. In a sense we may write the eigenvalue equation p | p p | p , i p x p p x, x which implies p x expipx / . Unfortunately we again find that

p| p

dx |expipx / | 2

diverges, so we are unable to enforce normalization of the momentum eigenstates. Nevertheless we shall see that they are quite useful, and pick the convenient normalization 1 expipx / . x|p 2

Here we should recognize that the wave functions (position representations) corresponding to momentum eigenstates themselves constitute a basis for the space of wave functions the Fourier basis! In particular, if we have a given wave-function x we know there is a function 1 dx expipx / x dx p | x x | p | , p 2 1 dp expipx / p. x x | dp x | p p | 2 (Hence x and p are related by Fourier transform.) In more familiar terminology, if x is the vector representation of a state | in the position basis then p is its representation in the momentum basis: p | p. The advantage of knowing p is that the action of the momentum operator is especially easy to express in this basis, e.g., 1 | p | 2 p

dp 1 | p p | 2 p
1 dp p p 2 p,

where again we have invoked the (infinite-dimensional) spectral decomposition,

dp | p p | p.

Note that we can already derive a commutation relation between the observables x and p. Working in position representation, 1 | x, p | 2 1 | xp | 2 1 | px | 2 i dx x x 2 x x x 2 x 1 1 x x i i

dx

x x x 2 x 1

x 2 x x x 1 1

2 x

dx x 2 x 1

i 1 | 2 .

Since this is true for arbitrary states | 1 , | 2 we must have at the operator level x, p i 1, or simply x, p i as it is more commonly written.

Recall the canonical commutation relation x, p i, where the scalar value on the right hand side implicit multiplies an identity operator on the relevant Hilbert space. Since these two operators have a nonzero commutator, we know that there exists an uncertainty relation x p 1 | x, p | . 2 2 That is, there is no state in the state space for which the product of the uncertainties in position and momentum is less than /2. Hence we are somehow forced to conclude that there are no states in the state space with vanishing uncerainty product not even position or momentum eigenstates. To try to see what is going on, let us first think about the uncertainties for a momentum eigenstate. Clearly we have p 0, but the wave function for a momentum eigenstate x | p p x expipx / is as uncertain in position as can be x , since | p x| 2 1 everywhere! Similarly for a position eigenstate x x x x , we have x 0 but the Fourier transform gives 1 dx expipx / x x x p 2 1 expipx / , 2 and | x p| 2 is a constant for all p. Hence p for a position eigenstate. One should be careful in thinking about the product of zero and infinity, but what the x p uncertainty relation really expresses is the following. By adding up Fourier components (momentum eigenstates) we ought to be able to make up wave functions with varying x and p, ranging from delta-functions x x on one extreme to plane waves expipx / on the other. In between we have things like Gaussian wave functions 2 1 x exp x , 1/4 2 x x p 1 2

dx expipx / x,

0 .7 0 .6 0 .5 0 .4 0 .3 0 .2 0 .1 0 -6 -4 -2 0 2 4 6

for which x 2 | x 2 | | x | 2 1 x 1 x

dx exp
1 x

x 2 2 x

2 x 2 exp x 2 x 2 x exp x 2 x 2

x 2 dx exp 2 x
2

2 dx x exp xx

1 x

2 dx x exp xx

x x 1 2 x

x , 2 and (recalling p i x ):

p 2

2 x

dx exp
i x

x 2 2 x

2 exp x 2 2 x x 2 exp x 2 2 x x x exp x 2 x 2 x x exp x 2 x 2 x 1 exp x 2 x 2 x 2 2 x x


2 2

x 2 dx exp 2 x x 2 2 x x

2 x

dx exp
i x

x 2 dx exp 2 x x 2 2 x x 2 x

2 x

dx exp

2 2 x 2 exp x 2 x x

2 x x

dx exp

dx x 2 exp

x 2 x

2 x x 2 x 2 2 x x

1 2 x 1 2 x 2 . 2 x (Note that we could have computed this using the Fourier transform instead...) Hence
2 xp x . 2 2 2 x

As we try to make the Gaussian narrower and narrower in x by reducing x , we see that the width of its Fourier transform grows as 1/ x and in just such a way that the product of uncertainties satisfies the Heisenberg bound! This coordination between x 0 and simultaneously p underlies the wave-particle duality in quantum mechanics, which well discuss further in the next lecture. From one point of view this is some kind of crazy magic in infinite-dimensional Hilbert spaces, but from another its really just a routine property of Fourier transforms.

We can now write down the Schrdinger Equation for a free particle, where by analogy with classical mechanics we have the free-particle Hamiltonian p2 H . 2m Hence p2 i d | t | t , 2m dt which we may rewrite as a wave equation by projecting onto the position basis:

i d x | t 1 x | p 2 | t , 2m dt x, t i ds s x 2 s, t 2m t s 2 x, t i 2 x, t. 2m x 2 t This is now a partial differential equation for the wave-function itself, corresponding to the set of coupled ordinary differential equations we would get in finite dimensions by projecting the Dirac/operator Schrdinger Equation onto some specific orthonormal basis for the Hilbert space: k | i d | H | d c k i k | H | m c m , dt dt
m

c k | k .
k x

Its interesting to note that we can make things that look like dimensional case by considering H of the form H 1 0 1 0 0 0 1 1 ,

in the finite

d c k c k c k1 c k , dt k where the last association is not meant to be rigorous! Then wed get something like H 1 0 2 1 1 2 0 1 ,

2 x 2

with

d c k c k1 2c k c k1 2 c k . dt k 2 Hopefully this helps to illustrate the idea that x or p in x, t or p, t corresponds to a continuous index for the wave-function, which itself corresponds to an infinite vector of complex coefficients for an orthonormal basis expansion. Also, differential operators that appear in the wave equation are linear, Hermitian operators! (Dont worry about the fact that our cheap and easy matrix version of x p doesnt look symmetric...)

Anda mungkin juga menyukai