Sie sind auf Seite 1von 10

Foundations of Physics, Vol. 37, No. 1, January 2007 ( 2007) DOI: 10.

1007/s10701-006-9089-1

Time-Symmetric Quantum Mechanics


K.B. Wharton1
Received April 9, 2006; revised October 12, 2006; Published online January 6, 2007 Communicated by Alwyn van der Merwe A time-symmetric formulation of nonrelativistic quantum mechanics is developed by applying two consecutive boundary conditions onto solutions of a timesymmetrized wave equation. From known probabilities in ordinary quantum mechanics, a time-symmetric parameter P0 is then derived that properly weights the likelihood of any complete sequence of measurement outcomes on a quantum system. The results appear to match standard quantum mechanics, but do so without requiring a time-asymmetric collapse of the wavefunction upon measurement, thereby realigning quantum mechanics with an important fundamental symmetry. KEY WORDS: quantum mechanics; time reversal symmetry; time reversal operator.

1. INTRODUCTION The Copenhagen Interpretation (CI) of quantum mechanics, while successful, is not time-symmetric. This asymmetry is somewhat evident in the Schrodinger equation itself, where substituting t t yields a different (though essentially similar) equation. A more dramatic timeasymmetry is found in the collapse postulate; upon measurement, a wavefunction collapses to a pure state only in the forward time direction. The time-reverse of this process is not permitted. For example, if atom A emits a photon that is absorbed by atom B, CI describes the photon as a spherical wave leaving A in all directions. According to CI, atom B itself (or perhaps a later measurement of atom B) collapses this spherical wave into a wave packet having a more precise trajectory, traveling from A to B. But applied to a lm of this same photon transaction
1

Department of Physics, San Jos State University, San Jos , CA 95192-0106, USA; e e e-mail: wharton@science.sjsu.edu 159
0015-9018/07/0100-0159/0 2007 Springer Science+Business Media, LLC

160

Wharton

played backwards, CI would be forced to conclude that the spherical wave was centered around B, not A, because it is impossible to uncollapse a wavefunction. This dramatic disagreement between CIs forward-time and reverse-time interpretation of the same physical event is evidence that this standard interpretation may not correspond to what is physically happening. In fact, one could argue that the collapse postulate of CI is the single most egregious example of time-asymmetry in all of modern physics. Still, it is notable that this is merely an asymmetry of the interpretation, not the physically observable results of quantum mechanics itself, which are as time-symmetric as the boundary conditions on a given system.(1) This fact alone provides sufcient motivation to search for an interpretation of quantum mechanics in which a forward-time and reversed-time perspective of the same physical events would be interpreted in the same manner. There have been several previous efforts to accomplish this task. Grifths,(2) Unruh,(3) Gell-Mann and Hartle(4) and Schulman(5) have effectively attempted to impose symmetrical boundary conditions (past and future) onto solutions of the Schrodinger Equation (SE). The previous work shows that it is difcult to nd general solutions to such a system. Indeed, one would hardly expect to nd an exact solution to the SE when imposing twice the usual number of boundary conditions. One solution to this dilemma would be to use the insight from Cramers Transactional Interpretation,(6) where the time-reversed SE was assumed to be on equal footing with the ordinary SE; this doubles the parameter space of the solutions in a natural way, and would therefore allow for a second boundary condition to be imposed. However, Cramer did not use symmetric boundary conditions and invoked a version of the collapse postulate that was as time-asymmetric as the one it replaced. To my knowledge, this paper is the rst attempt to combine the symmetric aspects of these previous proposals, applying two symmetric boundary conditions onto a time-symmetric version of the Schrodinger equation. The previous work most mathematically similar to this approach is Aharanov and Vaidmans Two State Vector formalism(7,8) (see, in particular, the last paragraph of Ref. 8), but this proposal will impose a boundary condition at every measurement (among numerous other differences). 2. A TIME-SYMMETRIC WAVE EQUATION In order to impose two consecutive boundary conditions on a quantum system and still nd general solutions, one must double the number of free parameters in the quantum system. It is therefore relevant that Schrodinger originally derived his equation from the non-relativistic limit

Time-Symmetric Quantum Mechanics

161

of the time-symmetric Klein-Gordon equation, and purposefully dropped one of the solutions in the process. Keeping both solutions would have resulted in both the SE and its time-reverse: iH|(t) = h |(t) , t |(t) . t (1)

iH|(t) = h

(2)

Here H is the Hamiltonian operator, |(t) is the solution to the SE and |(t) is the solution to the time-reversed SE; this notation will be used throughout to distinguish the two. To double the number of free parameters, these independent equations could be combined in some fashion. (This procedure is independently motivated by the original concern of imposing time-symmetry onto time-asymmetric models, and indeed a similar proposal has been made by Cramer(6) ). One way to combine these equations is to write a time-symmetric SE, iH 0 0 iH (t) (t) = h t (t) (t) = h |C(t) , t (3)

where |C is a wavefunction with double the usual dimensionality and has the generic solutions to (1) and (2) as its upper and lower components, respectively. The rst postulate of this approach is simply that a quantum system is best represented by the wavefunction |C from (3) (instead of the ordinary SE). Note that just because | and | solve conjugate equations does not at all imply that | = | ; instead |C has twice the free parameters of either | or | alone. One objection to such a postulate is the claim that | represents negative energy solutions, but this assumes the standard CI is correct, and therefore this criticism should not apply to a proposed replacement for CI itself. To see how | can represent positive energy solutions, consider the free wave solution (k, ) = exp(ik r it). This wave is traveling in the opposite direction from (k, ) = exp(ik r + it), but is essentially similar to (k, ) = exp(ik r+it). The precise equivalence is between (k, ) and (k, ); these two (different) terms should therefore correspond to a measurement of the same physical state; a free wave with momentum vector (The relative sign ambiguity is imporhk. tant, and will be discussed below.) It should be clear from this example that one cannot apply the ordinary momentum operator, P = i onto | ; the resulting momentum eigenh, values would be opposite the direction that the eigenfunctions were actually

162

Wharton

moving. This fact should come as no surprise; the relationship between (1) and (2) is that of a time-reversal, while standard quantum operators have been designed to apply only to (1). In order to continue to use the usual form of the momentum operator on | , one needs to rst time-reverse | using the antiunitary time-reversal operator T.(9) In other words, the ordinary momentum operator can still be applied to T| . Another problem is that a standard general operator Q cannot be applied to |C because they are of different dimensions. Instead of constructing a new measurement operator, a simpler way forward is to decompose |C into two wavevectors of the appropriate dimension. Two suitable wavevectors can be constructed from |C using a rectangular transformation matrix with two possible internal signs: |C = (I T) |C = | T| , (4)

where T is the time-reversal operator and I is the identity matrix, both with the same dimensionality as the ordinary Hamiltonian H. |C+ and |C are the simplest non-trivial representations of |C that allow for physically meaningful results when operated on by Q. It is natural to choose these two particular representations because the operator T does not have a welldened overall sign. (This holds for both bosons and fermions, which satisfy T2 = 1 and T2 = 1, respectively.) Normally this sign ambiguity is irrelevant, but in this case (and in the above free-wave example) it is unclear whether to add or subtract | and T| . To be consistent with the philosophy which led to the rst postulate, this approach will not eliminate one of the possibilities in (4), but instead will consider both |C+ and |C to be meaningful representations of |C and attempt to treat them on an equal footing. (Indeed, we must use both or else we will have lost information by compressing |C into a wavevector with half the dimensionality.) Because we are going to consider eigenvectors of the form a|n b|n , where a and b are complex constants, it is convenient to write the upper half of |C in the basis of the Hamiltonian eigenvectors |n , and also to dene the basis for the lower half of |C such that T|n = |n . (For the rest of this paper, |n represents the nth eigenvector of H with associated eigenvalue En .) In this basis, at any instant one can always write |C =
n

an |n bn |n

(5)

|C =
n

(an bn )|n ,

(6)

Time-Symmetric Quantum Mechanics

163

where an and bn are complex coefcients, and (6) has been simplied using T|n = |n and the antiunitary effect of T on complex numbers; Tbn = bn T. Note that the sum in (5) needs to only span the original dimensionality of H, not the doubled dimensionality of |C . 3. TWO-TIME BOUNDARY CONDITIONS Now we have reached the most difcult conceptual point in this approach: applying two consecutive boundary conditions to the total wavefunction |C . Although applying multiple spatial boundary conditions is ubiquitous is physics, our innate sense of causality often rebels at the idea of imposing future boundary conditions on a system. Nevertheless, for the sake of complete time-symmetry, we have no other choice. In what follows, the prior boundary will be imposed at a time t = 0, and the later boundary will be imposed at t = tf . These boundary conditions will presumably be related to measurements of the wavefunction at these two times. A specic relationship between measurements and boundary conditions is now proposed: Second Postulate: each measurement Q of a wavefunction (at some time t0 ) imposes the result of that measurement as an initial boundary condition on |C+ (t0 ) , and as a nal boundary condition on |C (t0 ) . In other words, instead of collapsing to an eigenvector upon measurement, the wavefunction is simply constrained to be that eigenvector! Because measurements perturb the wavefunction, |C will be discontinuous at t = t0 ; therefore the distinction between initial and nal boundary conditions is important. An initial boundary condition governs the behavior of the wavefunction into the future of the measurement; this applies to |C(t0 ) after any measurement-induced discontinuity. We can specify that the boundary condition applies to the positive-time side of the disconti+ nuity by using the notation |C(t0 ) . Similarly, the nal boundary condi tion is applied to |C(t0 ) , just before any discontinuity occurs. While this postulate may seem to distinguish past from future, it is at least antisymmetric, and exchanging |C+ |C (or equivalently T T) would not change any of the following results. Using this notation, the boundary conditions proposed by the second postulate can be expressed as the pair of eigenvalue equations (with measured eigenvalue qn )
Q|C (t0 ) = qn |C (t0 ) .

(7)

Even though the solutions to (3) now have twice as many boundary conditions as a wavefunction in standard quantum mechanics, they also

164

Wharton

have twice the number of free parameters. This means that solving for | and | will generally be possible. Also, it means that there will be solutions that evolve directly from one measurement to the next without requiring a collapse to arrive at the second measurement. This is because the wavefunction |C(t) for times 0 < t < tf already takes into account the boundary condition at t = tf . In other words, the components | and | depend on the next measurement that will be made (and the result of that measurement), and their values must therefore conform to that future boundary condition. External causality is safe, because in order to get information about the future boundary one would need to measure the wavefunction, which by denition could not happen until t = tf . Furthermore, the additional parameters in |C are not ruled out by Bells inequality,(10) because their dependence on the future measurement means that they are not strictly local hidden parameters as dened by Bell. 4. TIME-SYMMETRIC PROBABILITY We now can examine the case in which the Hamiltonian, H, is timeindependent. For simplicity, also assume H has discrete eigenfunctions, and that all measurements are complete (meaning a maximum possible number of outcomes). At t = 0, suppose some measurement Q is made, and nds the wavefunction in a pure eigenstate. From the second postulate, we associate this measurement with an initial boundary condition on |C+ (0+ ) . (It is also a nal boundary condition on |C (0 ) , but because of the measurement-induced discontinuity, this need not affect anything that happens after t = 0.) Standard quantum mechanics would say that the prepared wavefunction could be written as a sum of cn |n terms, where cn are known complex coefcients, determined by the original measurement Q. In this new approach, we instead have |C+ (0+ ) =
n

cn |n =
n

(an + bn )|n , an |n bn |n .

(8) (9)

|C(0+ ) =
n

Here an and bn are unknown complex coefcients; only the sum (an + bn ) = cn is well-dened by the initial boundary condition, not an or bn individually. This is because |n and T|n represent the same physical state (they have been combined in (8)), and a single measurement cannot resolve their distinct coefcients.

Time-Symmetric Quantum Mechanics

165

Because we are in the Hamiltonian basis, (9) can easily be timeevolved using (3), yielding |C(t) =
n

an exp(iEn t/ n h)| h)| bn exp(iEn t/ n

(10)

Now we can construct |C and impose the nal boundary condition at t = tf . Because of the relative minus sign that appears between the an and bn coefcients in |C , one nds that |C is not determined by the initial measurement. Imposing a second boundary condition of a different measurement Q in ordinary quantum mechanics would collapse the wavefunction into some state that is equal to a sum of dn |n terms, with known complex coefcients dn that are determined by this later measurement. But as proposed by the second postulate, in this case there is no collapse, but instead an exact equivalence:
|C (tf ) = n

dn |n =
n

(an bn ) exp(iEn tf / n , h)|

(11)

where the sign of the exponent in the lower half of |C is reversed upon operation by T. No collapse is necessary for |C to become a pure eigenstate of Q because this measurement has been imposed as a boundary condition on |C itself. This continues to be true for the next measurement at, say, t = tf , so long as the measurement at tf sufciently perturbs the wavefunction in such a way that (8) no longer applies for t > tf , and (11) thereby becomes the initial boundary condition on |C+ for the next timespan, tf < t < tf . While it is true that the discontinuity at each measurement might be interpreted as a new sort of collapse, the important point is that this discontinuity would be described in the same manner in both a forward-time and backward-time perspective, in sharp contrast to the fundamentally time-asymmetric Copenhagen Interpretation. Using the initial and nal boundary conditions, one can now solve for an and bn by extracting the coefcients from (8) and (11): an + bn = cn , (an bn ) exp(iEn tf / = dn . h) (12)

Solving these equations for an and bn and then inserting these results into (10) gives us a known wavefunction |C at all times between the two measurements. But this can only be done after the nal measurement is madewhen the values of dn become known. Of course, this hindsight is useless for predictive purposes. We therefore need to construct a function

166

Wharton

P that yields the likelihood of any consecutive pair of measurements at times 0 and tf . To be consistent with the overall symmetry behind this approach, P should also be symmetric under C+ C , an bn , and t = 0 t = tf . Then, for any known values of cn and potential values of dn , we could calculate P to predict the probability of making that particular nal measurement. The correct answer is known from standard quantum mechanics (assuming standard normalization) to be
2 2

P =
n

dn cn eiEn tf /h

=
n

(an

bn )(an

+ bn ) ,

(13)

where the c and d coefcients have been substituted from (12), naturally canceling the exponential terms. This can be written in the manifestly time-symmetric form
P = C (0+ )|C+ (0+ ) C+ (tf )|C (tf ) .

(14)

The superscripts indicate that the wavefunctions in (14) need to be evaluated after any discontinuity caused by the initial measurement, and before any discontinuity caused by the nal measurement. It is promising that the positive probability P can be expressed in this simple manner, obeying all of the necessary symmetries. Extrapolating from (14), the relative likelihood of any complete sequence of measurements on a wavefunction |C , at times t1 , t2 , . . . , tN is therefore
N1

P0 =
n=1

+ + C (tn )|C+ (tn ) C+ (tn+1 )|C (tn+1 ) ,

(15)

where N > 1 and each measurement is constrained by the boundary conditions from (7). Given a third and nal postulate that the probability ratio of any two measurement sequences A and B is the ratio P0 (A)/P0 (B), this completes the demonstration of a full equivalence between this approach and standard quantum mechanics for a discreteeigenvalued and time-independent Hamiltonian. The picture described here is that the wavefunction |C selects a solution that conforms to all of its boundary conditions from a space weighted by P0 . Because some sequences of boundary conditions have a larger number of possible solutions (that number being proportional to P0 ), those measurement sequences will appear to be more probable. In the laboratory, of course, one is interested in the relative probabilities of the

Time-Symmetric Quantum Mechanics

167

nal measurement in a chain, but in principle this third postulate would apply to any sequence of known and unknown measurements. Substantial work remains to be done to show that this approach will work for incomplete measurements, as there would not be sufcient information to fully determine |C from two consecutive measurements if one of them was not complete. 5. IMPLICATIONS AND SUMMARY One important question is whether this approach might yield different predictions from the Copenhagen Interpretation, providing a way to experimentally distinguish the two pictures. While there is no clear answer to this question, the most promising possibility lies in the second postulate, in that each measurement imposes boundary conditions both before and after the measurement-induced discontinuity. This might imply that in a situation where multiple measurements were made at the same time (to within some fundamental uncertainty), this time-symmetric approach could very well yield novel predictions. This possibility will be addressed in a future publication. Even if it turns out that there are no new predictions of this timesymmetric approach, it still has some novel theoretical aspects (although ease of calculation is not one of them). Most notably, by removing the need for a time-asymmetric collapse, many open issues regarding quantum measurement might be given a more satisfactory explanation, ideally helping to answer still-open questions concerning why quantum mechanics has to be the way that it is. For example, the square in the standard probability measure in (13) might turn out to be a natural consequence of simply requiring (14) and (15) to have a time-symmetric form. This approach also realigns quantum mechanics with an important fundamental symmetry. However, it certainly remains possible that this interpretation might raise more questions than it answers. To summarize, I have proposed changing the Copenhagen Interpretation of quantum mechanics in three substantive ways, each corresponding to a postulate stated above. (1) The wavefunction is no longer a solution of the Schrodinger Equation, but instead is |C(t) , a solution to the timesymmetric Eq. (3). (2) Instead of a collapse postulate, this formulation imposes a boundary condition on the wavefunction at every measurement, equal to the outcome of that measurement. Each measurement is an initial boundary condition on one representation of the wavefunction, |C+ , and a nal boundary condition on another representation, |C . (3) Instead of the standard probability formula, the relative probability of any complete

168

Wharton

measurement sequence is proportional to the time-symmetric function P0 in (15). While these results need to be generalized, they strongly imply that it is indeed possible to have a fully time-symmetric formulation of quantum mechanics, without requiring a time-asymmetric collapse of the wavefunction upon measurement. ACKNOWLEDGEMENTS The author gratefully thanks J. Finkelstein for his unfailing insight over the course of our numerous discussions on this research. Further thanks go to W. Wharton and L. Schulman for feedback and guidance. REFERENCES
1. Y. Aharonov, P. Bergmann, and J. Lebovitz, Time symmetry in the quantum process of measurement, Phys. Rev. 134, B1410 (1964). 2. R. B. Grifths,Consistent histories and the interpretation of quantum mechanics, J. Stat. Phys. 36, 219 (1984). 3. W. G. Unruh, Quantum measurement, in New Techniques and Ideas in Quantum Measurement Theory, D. M. Greenberg, ed. (New York Academy of Science, New York, 1986) p. 242. 4. M. Gell-Mann and J. B. Hartle, Time symmetry and asymmetry in quantum mechanics and quantum cosmology, in Proceedings of the NATO Workshop on the Physical Origins of Time Asymmetry, J. J. Halliwell, J. Perez-Mercader, and W. Zurek, eds. (Cambridge University Press, Cambridge, 1994). 5. L. S. Schulman, Times Arrows and Quantum Measurement (Cambridge University Press, Cambridge, 1997). 6. J. G. Cramer, Generalized absorber theory and the Einstein-Podolsky-Rosen paradox, Phys. Rev. D 22, 362 (1980); The transactional interpretation of quantum mechanics, Rev. Mod. Phys. 58, 647 (1986). 7. Y. Aharonov and L. Vaidman, Complete description of a quantum system at a given time, J. Phys. A 24, 2315 (1991). 8. Y. Aharonov and L. Vaidman, The two-state vector formalism of quantum mechanics, in Time in Quantum Mechanics, J. G. Muga et al., eds. (Springer, Berlin, 2002), p. 369. 9. E. P. Wigner, On the time inversion operation in quantum mechanics, G ttinger o Nachr. 31, 546 (1932); E. P. Wigner, Group Theory (Academic, New York, 1959), Chap. 26. 10. J. S. Bell, On the problem of hidden variables in quantum mechanics, Rev. Mod. Phys. 38, 447 (1996).

Das könnte Ihnen auch gefallen