2.

1
A beam of spin 1 atoms goes through a series of Stern2 Gerlachtype measurements as follows: a. The rst measurement accepts sz = h/2 atoms and rejects sz = /2 atoms. h

b. The second measurement accepts sn = h/2 atoms and rejects sn = /2 atoms, where sn is the eigenvalue of the operator h S n, with n making an angle in the xz-plane with respect to the z-axis. c. The third measurement accepts sz = /2 atoms and rejects h sz = h/2 atoms. What is the intensity of the nal sz = /2 beam when the h sz = h/2 beam surviving the rst measurement is normalized to unity? How must we orient the second measuring apparatus if we are to maximize the intensity of the nal sz = /2 beam? h (Sakurai, problem 1.13.) The state after the rst measurement can here be regarded as the initial state, since it is normalised. That means that we are starting out with |+. The ltering process can be described as |+ = a |+n + b |n a |+n = c |+ + d | d | . This ltering can be carried out using projection operators: d | = | | |+n +n | |+ = | |+n +n |+ In the end we are after the nal intensity |d|2 : 2 2 |d|2 = |+n +n |+ (1)

In order to compute this, we need the eigenstate |+n of the operator S n: sin S n = Sx Sy Sz 0 = Sx sin + Sz cos = cos h h . = |+ | + | +| sin + |+ +| | | cos = 2 2 . h cos sin = 2 sin cos 1

We can quickly establish the eigenvalues /2. We can nd the eigenvectors h with Gaussian elimination, but to do that we would really like to be able to multiply and divide with abandon we need to consider a few special cases rst, to keep the trigonometric functions away from the dangerous values 0 and 1. Specically, we need to treat {0, /2, } separately. For = 0, S n = Sz and |+n = |+, |n = |. 1 For = /2, S n = Sx and |n = 2 (|+ |). For = , S n = Sz and |+n = |, |n = |+. Now, for any other : sin . h cos 1 S n 1I = sin cos 1 2 sin sin 1 1 cos 1 cos 1 sin cos 1 0 0 . x1 If we take |+n = , this means that x1 and x2 obey the equation x2 (1 cos ) x1 = sin x2 . Using the formulae for twice an angle, we can simplify things a bit: x1 = 2 sin cos x2 2 2 2 = sin x1 = cos x2 2 2 Combine this with the normalisation condition |x1 |2 + |x2 |2 and we immediately see a worthy candidate: . cos 2 |+n = cos |+ + sin | = . (2) sin 2 2 2 2 sin2 This general expression also reproduces the special cases isolated at the outset, so we may take (2) to be the desired equation for |+n . There is no need to calculate the other eigenstate |n . Going back to the expression (1) for the nal intensity, we see that we need the inner products: +|+n = cos 2 and |+n = sin 2

1 sin2 = sin2 . 2 2 4 So, if the incoming intensity was unity, the nal intensity will be 1 sin2 . 4 To maximise the nal intensity, the second measuring apparatus must be set at right angles with the rst: = /2 produces the maximal intensity 1 . 4 |d|2 = cos2 2

2.2
# -*-octave-*nn = 60; id = eye ( nn ); h = - ( shift( id, 1) + shift( id, -1 ) ); h(1,1) += 3; [u,e] = schur(h,"a"); ev = diag( e ); shouldbezero = h - ( u * ( e * u ) ); deltat = .5; ut = u * diag( exp( i * deltat * ev ) )* u ; x = psi psi psi [0:nn-1]; = exp( -( (( x - nn/2)/(nn/6) ).^2)); = psi ./ sqrt( psi * psi); = ( exp( 32i * pi * x / nn ) .* psi ).;

gset key off for t = 1:120 axis([1 nn 0 .1]); plot( abs(psi).^2 ) axis([1 nn 0 .1]); psi = ut * psi; eval(sprintf("print(\"-dpng\", \"fil%.3d.png\")", t )); end system("convert fil???.png gr.gif"); Here comes every fth frame in the time evolution:

2.3
Two observables A1 and A2 , which do not involve time explicitly, are known not to commute, [A1 , A2 ] = 0, yet we also know that A1 and A2 both commute with the Hamiltonian: [A1 , H] = 0, [A2 , H] = 0. Prove that the energy eigenstates are, in general, degenerate. Are there exceptions? As an example, you may think of the central6

force problem H = p2 /2m + V (r), with A1 Lz , A2 Lx . (Sakurai, problem 1.17.) Since [A1 , H] = 0, we can diagonalise them simultaneously. Let the basis that diagonalises A1 in this way be denoted by {|a(i) }. Similarly, since [A2 , H] = 0 we can diagonalise A2 and H simultaneously; let that basis be denoted by {|b(i) }. Since [A1 , A2 ] = 0 we know that, in general, |a(i) does not equal any of the |b(i) . It will be some linear combination: (j) |a(i) = ci |b(j) .
j

For at least one i we will have several nonzero ci otherwise we could simultaneously diagonalise A1 and A2 , and then we would have [A1 , A2 ] = 0. (i) (i) Since |a(i) is an eigenket of H, we have H |a(i) = ha |a(i) where ha is (j) the energy eigenvalue of the eigenstate |a(i) . Similarly, let hb be the energy eigenvalue of the eigenstate |b(j) . Now, on the one hand we have (j) (j) (j) H |a(i) = ci H |b(j) = ci hb |b(j) ,
j j

(j)

and on the other hand we have

H |a(i) = h(i) |a(i) = a which means that

j

ci h(i) |b(j) , a

(j)

ci hb |b(j) =

(j) (j)

Since the |b are linearly independent, we can equate their coecients. (j) (i) (j) Thus, provided that ci = 0 we have ha = hb . But, as pointed out above, (j) there must be at least some i for which ci = 0 holds for several j, which (j) means that for this i we have several hb taking on the same value. In other words: the energy eigenstates of H are degenerate.

(j)

ci h(i) |b(j) . a

(j)

2.4
Prove the operator identity F pi , F (x) = i h xi where F (x) is a function of the operator x in two ways: (1) directly from the commutation relations [xi , pj ] = i ij and (2) by using h the real space basis and evaluating |[pi , F (x)]|f . x 7

Using the commutation relations. Assume that the operator F (x) can be written in a Taylor expansion: F (x) = a,, x x x . 1 2 3
,,

Without loss of generality, assume i = 1. (If it is not, we relabel the axes so that it is.) We then compute [p1 , F (x)] = a,, [p1 , x ]x x . 1 2 3
,,

Noting that [a, b] = [b, a], we can now call into eect the theorem of last weeks homework, number 1.3: [[, ], ] = 0 = [m , ] = mm1 [, ] for operators , . In this problem, we take = x1 and = p1 . [x1 , p1 ] = i , which commutes h with x1 , so the requirements are fullled. Thus [p1 , x ] = x1 [p1 , x1 ] = i x1 [p1 , x1 ] = i h 1 h 1 1 [p1 , F (x)] = i h F x x x = i h 1 2 3 x1 x1 ,, x 1 x1

If we now undo the relabelling that resulted in i = 1, we arrive at the statement to be proved: [pi , F (x)] = i h F . xi

Using the real space basis. Ill take x to be the operator, and x to be the coordinates of the state | . I would have found it more intuitive to have x it the other way around. . . In general, we have |f (x)| = dx (x )f (x ) (x ). |[pi , F (x)]|f = |pi F (x)|f |F (x)pi |f x x x |pi F (x)|f = x dx ((x x)) (i ) h 8

F (x )f (x ) = xi

This gives us the following expression for the left hand side: F (x ) f ( ) x |[pi , F (x)]|f = i x h x
i x = x

F (x ) f (x ) = i dx (x x) h f (x ) + F (x ) xi x i F (x ) f (x ) = i h f ( ) + F ( ) x x xi x = xi x = x x f (x ) |F (x)pi |f = i F ( ) x h x x
i x = x

Now on to the right hand side: F F (x ) | i x h |f = dx ((x x)) i h f (x ) = xi xi F (x ) = i h f ( ) x xi x = x We see that the expression for the left hand side above and this expression for the right hand side agree. Thus [pi , F (x)] = i h F . xi

2.5
Let the state initial state [sic] of a one dimensional free particle be given by |(0) where x|(0) is given by the Gaussian 1 2a 4 ax2 x|(0) = e

where a is real and positive. (a) Transform to momentum eigenstates and show that 1 4 p2 1 h p|(0) = e 4 2 a 2a 2 h (b) Compute x|(t) and show it is equal to 1 ax2 /(1+(2i at/m)) h 2a 4 e 1 + (2i at/m) h

(c) Compute the probability density | x|(t) |2 and show that it is a gaussian [sic] which spreads out in time. 9

Transformation to momentum eigenstates proceeds as usual p|(0) = dx p|x x|(0) We know that 1 x|p = exp 2 h ipx h ,

so

where the last step was taken using the familiar expression for Gaussian integrals with an imaginary coecient for the x factor. Cleaning this up, we arrive at 1 4 1 p2 p|(0) = exp 2 2 a h 4a 2 h as desired.

1 4 1 ipx 2a p|(0) = dx exp exp(ax2 ) = h 2 h 1 1 2a 4 ipx 2 = dx exp ax = h 2 h 1 1 1 2a 4 2 p 2 = exp /4a a h 2 h

To compute x|(t) we need the time evolution operator U : U = exp(iHt/ ), h where H = p2 /2m for a free particle. We then have x|(t) = x|U |(0) = dp x|U |p p|(0) =

ip2 t = dp exp x|p p|(0) = 2m h 1 4 1 1 ip2 t ipx p2 = dp exp + = 2m h h h2 4a 2 h 2 2 a h 1 4 1 1 1 1 it ix 2 = dp exp + p + p = 2 2a 2 m h h h2 h 2 2 a h 10

1 = 2 h

1 2 2 a h

1 4

2 1 + 2a 2 h

it m h

1 2

exp

1 it 1 + 2i at/m h = . 2 + m h 2a h 2a 2 h This clears up the exponential nicely, and with some further thought we nd the coecient matches up as well: x|(t) = 2a 1 4 exp (ax2 /(1 + 2i at/m)) h 1 + 2i at/m h

Note that

x 2 h

1 it 2 + h 2a 2 m h

As far as I can see, there is nothing in the Gaussian formula used for this that tells us which branch of the complex square root we should use Im not very happy with the derivations of this formula I have seen. But assuming that it is correct, this is what we get. The probability density | x|(t) |2 is relatively easy. If c is a complex number, |c|2 = |c2 | = c c. We will use the rst of these equalities for the square root, squaring it rst, while we use the second for the exponential: 1 2 2a 4 exp (ax2 /(1 2i at/m)) exp (ax2 /(1 + 2i at/m)) h h = = 2 1 + 2i at/m h 1 1 exp ax2 (12i at/m) + (1+2i at/m) h h 2a = = 2 2 2 2 1 + 4 a t /m h = 2ax2 exp 1+4 2 a2 t2 /m2 2a h 1 + 4 2 a2 t2 /m2 h | x|(t) |2 =

This is a Gaussian (the only x is a x2 with a negative coecient, appearing in the argument of the exponential function), and it is spreading out in time. To see the latter, consider that the square of the standard deviation is proportional to the expression in the denominator, 2 1 + 4 2 a2 t2 /m2 , h and this increases with time. 11