Anda di halaman 1dari 8

Ph125b lecture notes, 1/04/01

Complementarity and Heisenberg s Uncertainty Principle

[Feynman] Figure 1-3: An experiment with electrons

4. Complementarity, classic illustrations of (recoiling slits, Feynman microscope)

[Feynman] Figure 1-4: A different electron experiment

If we let d be the separation between slit 1 and slit 2, the requirement on photon wavelength should be d to reliably distinguish the region of origin. In this case each photon will carry a momentum h 2 h. k photon p photon 2 d If we are imaging near the diffraction limit, the electron recoil p electron will be randomly distributed between roughly 0 and 2p photon .

Hence we see that in trying to localize the position of the electron to within a distance x d, we necessarily impart a momentum uncertainty p h/ x. What effect will this have
on the accumulating fringe pattern? Let L be the distance between the collision region and the final screen. We expect that the random displacement of the fringe pattern should be about p electron L h x fringes L uncertainty L p electron p electron d , whereas the fringe spacing should be L h. f
electron L p electron d d Hence we can expect that the fringes are washed out! Uncertainty principle? Field quantization, wavelength-momentum relation for photons, resolution limit ...

[Feynman] Figure 1-6: Feynmans recoiling slits

Here if the electron passes through slit 1, we expect the wall to recoil by p 1 d p electron, L and if the electron passes through slit 2, d p electron. p2 L

In order for the difference between these two to be discernible in a measurement of the final wall momentum, we need the initial momentum uncertainty for the wall to be p dp . wall L electron But if this is the case, then the walls initial position uncertainty must satisfy 1 x d wall p electron L 1 d electron L f. Hence the center of the fringe pattern (as determined by the midpoint between the two slits in the wall) has a greater uncertainty than the fringe spacing the pattern is washed out!

5. Is it always the Uncertainty Principle that enforces Complementarity? recoils from infinite-dimensional probes... finite-dim?

6. Bra-ket model with two intermediate states and subsequent evolution, interference

Suppose we block slit 2 before shooting an electron at the wall. The part of the wave function that leaks through slit 1 will be represented by some well-defined wave packet 1 x
| 1  . We leave aside the question of normalization but suggest that we can think of the wall as performing a measurement with two possible outcomes... Similarly,  2 x
| 2  if we block slit 1. Then if neither slit is blocked, we expect from superposition | wall    1 | 1  | 2 
2 just to the right of the wall. Note that we expect  | 1 2  0   since 1 x
and 2 x
should have negligible overlap. Free evolution (spreading) of the superposed wave packets can be represented by a unitary transform, | screen    U | wall  . Then the final probability to find the electron at a position x on the final screen is

Pr x, 1&2 | x | screen |2 1 | x | U | 1 x | U | 2 |2 2 1 | x | U | 1 |2 1 | x | U | 2 |2 2 2 | U Re 1
| x x | U | 2 Pr x, 1 Pr x, 2 Re | U
| x x | U | . 1 1 2  2 We recognize the rightmost term on the RHS as the interference term responsible for fringes. Its there because | wall is a coherent superposition of | 1 and | 2 . What if we had an incoherent superposition instead? In particular,  1 | 1 1 | | 2 2 | wall 2 leads to  1 U | 1 1 |U
U | 2 2 |U
, screen 2 and the probability distribution at the screen is Pr x, incoherent Tr | x x |  screen

Tr | x x | U | 2 Pr x, 1 Pr x, 2 . 1 2
1 2
1

Tr | x x | U |

|U

|U

Hence, as we should expect, no fringes.

Lets say we continue with this abstract line of reasoning, and represent our attempts to measure the particle trajectory (through slit 1 or 2) in the following way. Define a quantum meter system that lives in a two-dimensional Hilbert space, | M  H 2 . In contrast to the discretized representation above of the electron state (via two Dirac kets . Lets choose two orthonormal in H  ), imagine that this is really something like a spin- 1 2 basis states for H 2 and call them  |1  , |2   . Suppose the initial state of the meter is prepared as |1  . Our measurement interaction (assumed to take place at the wall) is M  1 M  |  1   1 |  |2  1 |  |1  2 |   |  2   2 |. Just to make sure this is implementable via some Hamiltonian, we check that M is unitary (note that it is clearly Hermitian): M  M MM ! MM

( to denote the identity operator on the two-dimensional subspace Here we are using 1 (of the electron Hilbert space H ) ) spanned by * | + 1 , , | + 2 ,- . So the effect of the measurement interaction is to map the joint state according to
electron

" | # 1 $&% # 1 | ' ( electron . 1M " 1


1M

1M

" | # 2 $&% #

1 |1 | 1 |1 | 2
2 1 |1 | 1 |2 | 2 
. 2 Hence we establish entanglement between the electron and the meter system. If we want to know which way the  electron went, we could simply perform a measurement of the meter system in the  |1 , |2 basis. Of course, if we do this before the electron reaches the screen, we know that the electron state will immediately be projected into either | 1 or | 2 and no fringes will be observed. But what if we wait until after? As far as the electron and its (possible) fringes are concerned, measuring after is the same as not measuring at all. In the latter case, we know that the proper way to proceed is to take
a partial trace over the meter states before computing the probability distribution Pr x . In the familiar old way from last term, we start by forming the overall density matrix 1  
 
 1| |1 | 1 |2 | 2 1 | 2| 2 | wall 2 |1 |
wall

1 2

 

|1 1 | |2 1 |

 

|1 2 | |2 2 | 2|

 

 

| |

then take the partial trace:

| 1 1 | | 2 2 | . 2 Here we recognize the incoherent superposition which does not produce fringes!

   1 |  1

wall |1

 



wall |2



7. Complementarity: if you try to distinguish alternatives by any means, they no longer interfere. Note that actually knowing the outcome of this measurement is not necessary, just the entanglement suffices... no necessity of recoils or implication of the Uncertainty Principle.

8. Distinction between UP and inference-disturbance principle(s)

Suppose we were to use a weaker measurement interaction like M   1M  |  1     1|

%) ' ( 1|" & , * + sin $ |1 % & cos $ |2 %'-( 2 | '/. | 0 2 %( 0 2 |. |1 % & sin $ |2 $  1 2 /2 this is equivalent to the original measurement, but if " $ 3 0 this is the (restricted) If " identity. For small $ , the effect of this interaction is to produce a weakly-entangled state % 1 1 * |1 % . | 0 1 %& |1 % . | 0 2 %' |1 % . | 0 wall  2 4 1 * |1 %. | 0 1 %&"* cos $ |1 %& sin $ |2  %/ ' . |0 2  %' 2 1 1 * |1 %.5* | 0 1 %& cos $ | 0 2 %'6& sin $ |2 % . |0 2  %'. 2
5

"!#! cos $

If we now trace out the meter, |1 1 | |

wall 1

cos |
1 

2 

1 2 

|  cos sin
 


2

2 |

1 2

 

|1  2 | 


 |

cos |
2     1

|2   1 |  sin |

|  cos
2  2

2 |

sin 2 |2  2 |  |

1 | 1  cos | 2 1 sin 2 | 2 2 |  1 |  cos  2 |    2 2  1  | 1 1 |  | 2 2 |  cos | 1 2 |  | 2 1 | .       2 Comparing this with the fully incoherent case, we see that 0 ! #"%$ /2 leaves a small coherence term in the reduced density matrix, which leads to fringes. At the same time, if we look at the joint quantum state just after the measurement interaction, we see that a projective measurement on the meter would only give us partial information on the whereabouts of the electron! If we imagine that is very small for example, and keep terms only to first order, 1 - |1 '.( - | * 1 '
/ | * 2 '01/32 |2 4.5 | 6 2 47 . M & |1 ')( | * wall , ' + 2 Then the probabilities and post-measurement states are clearly 1 8 | 6 1 4=< | 6 2 47 Pr 8 1 7:9 1, electron ; 2 1 Pr 8 2 7:9 electron ; | 6 2 4 . 2 , 2 In the small fraction of realizations in which we get outcome 2, everything is great. But most of the time we will get outcome 1 and we effectively learn nothing about which slit the electron went through!

 

So apparently, the mechanics of entanglement, measurement, and decoherence enforce some kind of tradeoff between how much we learn about the electrons trajectory, and how much we wipe out the fringes the more we infer the more we disturb. As mentioned early on in the first term of this class, however, we still dont know how rigorously to quantify this notion! Unlike the Uncertainty Principle, which is a constraint on the types of states that exist in a Hilbert space, these sorts of inference-disturbance relations are some kind of statement about the association between measurement and decoherence...

9. Spinor waves and entanglement, basic idea of Rempe paper: Use an atom instead of an electron, meter degree of freedom is actually an internal degree of freedom of the atom. [S. Durr et al, Nature 395, 33 (1998)]

Anda mungkin juga menyukai