Thursday, January 19, 2017

Quantum theory of the past



Eternalism is not hard to justify.
Yesterday I contemplated my situation and concluded: "This is real, this is now."

Today, when I recollect that scene, I'm inclined to think it was indeed real, and shows the reality of the past (which has not flickered out of existence but is .. elsewhere).

Yesterday, I also thought, "Tomorrow, I will be writing a post."

Today, here I am doing it. For yesterday's me, that shows the reality of the future.
As Frank Sinatra observed, "You can't have one without the other."

---

For greater conviction, we can appeal to special relativity. As I wrote in a piece for sciencefiction.com,
"Brian Greene in ‘The Fabric of the Cosmos’ (page 134) considers an alien in a galaxy ten billion light years away, at the edge of the visible universe. Simply by ambulating towards or away from us at 10 mph, the alien’s view of what is happening ‘right now’ on earth swings from 149 years in the past to 149 years in the future."
Still, the block universe is classical and therefore inaccurate. It's necessary to move to quantum theory where, as usual, one needs to take the red pill.

Theoretical physicist Jeremy Bernstein on 'A Quantum Past'.
"In FAPP (For All Practical Purposes) language we have a quantum mechanical system described by a wave function ψ(t), I am  only interested in the time variable.

The wave function obeys a Schrödinger equation with a Hamiltonian H. The formal solution to this equation is ψ(t) = exp(iHt)ψ(0). Throughout I am setting ћ = 1. Thus to recover Ψ(0) from ψ(t) all we have to do is to multiply by exp(-iHt).

Haven’t we then recovered the past? What is all the fuss about? The problem is that there is more to life than the wave function. There are the “observables” which represent what we really want to know about the system. These observables are described by Hermitian operators A. B. C and so on. We can expand ψ in a sum over the orthonormal eigenfunctions of any of these operators. The coefficients in the expansion are related to the probabilities that in a measurement the system will be found to have one of these eigenvalues. This is “Born’s rule” and in FAPP it must be assumed.

To find which of these eigenvalues the system actually has, we must perform a measurement. Stripped to its essence the apparatus that produces this measurement projects out from the sum of eigenfunctions one of them.

After the measurement the rest of the terms in the sum disappear. Using the term of art, the wave function “collapses”. It is at this point that we lose our capacity to reconstruct the past.

Projection operators are singular. They do not have inverses. All the king’s horses and all the king’s men cannot put the wave function back together again.

It was von Neumann in the early 1930’s who first noted that in FAPP mechanics there were two kinds of processes. There were processes that could be described by a Schrödinger equation and there were measurements which could not. He did not, as far as I know, comment on what this implied for retrodiction.

A case in point is an electron described by a spherically symmetric Schrödinger wave. If this electron strikes a detector is does so at a place - a spot. After this happens all trace of the spherically symmetric wave function vanishes.

I have certainly not made a careful search of the literature but among the founding fathers of FAPP I can come up with only two references that deal with the matter of the quantum past.

One is Heisenberg and the other is a paper by Einstein, Richard Tolman, and Boris Podolsky, “Knowledge of Past and Future in Quantum Mechanics” which they wrote in 1931 when Einstein was spending time at CalTech."
Bernstein talks about these two references, and then discusses where he thinks the problem resides, and what is to be done.
"It seems to me that any interpretation of the quantum theory that addresses this [the problem of wavefunction collapse] must have the feature that measurements are simply just another interaction like the rest.

Von Neumann’s notion that there were two classes of interactions one whose time evolution could be described by a Schrödinger equation and one of which couldn’t, has to be abandoned.

I will discuss two proposals for doing this each of which has its adherents and its detractors. On the one hand I am going to discuss what I will call “Bohmian mechanics” a term which David Bohm, who invented this approach , apparently did not like. As far as he was concerned, he was just doing quantum mechanics but in a different way. However nearly everyone else calls it Bohmian mechanics - so will I.

On the other hand, I am going to discuss the “decoherent history” interpretation which Murray Gell-Mann and Jim Hartle have done the most on. Sometimes this is called the “many worlds” interpretation, but not by them. I think that the term “many worlds” is misleading. As far as we know there is one world, the one we live in."
I'm not a fan of “Bohmian mechanics” and insofar as any quantum ontology works for me, it has to be "Many Worlds" (Sean Carroll explains why).

So I'll skip over Bernstein's description of Bohm's view of the quantum past and quote his take on “decoherent histories”.
"In the "Many Histories Interpretation" what indeed is history?

At first sight this might seem to be obvious. All we have to do is to run the chain backwards.

Yes this gives one history but there are others, possibly very many others. The reason is that if all we know is the present state vector there are many paths by which we could have arrived there depending on which initial state vector we started from. We have no way of knowing this from the data we have at hand.

Let us take an example discussed by Hartle - the Schrödinger cat (I can’t resist noting that when I spent an afternoon with Schrödinger in his apartment in Vienna there was no cat). In any event this unfortunate feline is put in a box that contains a capsule of poison gas and a sample of uranium. The capsule is triggered so that if the uranium has an alpha decay, the alpha sets off the trigger and the unfortunate feline expires.

After a time interval we open the box and happily the cat is alive. It could, according to the many history approach have arrived at this state in two ways. The initial state might have been a cat alive state or it might have been a coherent sum of a cat alive and a cat dead state. From the presence of the living cat we cannot decide.

The vision of the past given by the decoherent history interpretation and the Bohmian seems radically different. In Bohmian mechanics we could in principle follow all the cat molecules backwards in time and arrive at one and only one past.

I don’t know how you feel, but the ambiguity of the past makes me queasy. It might be entertaining to imagine that in an alternate past my grandmother who was born in a Polish stetl could have been Eleanor Roosevelt.

I readily accept that these pasts to not communicate but there seem to be too many of them from the point of view of economy. A trip to a barber wielding Occam’s razor seems warranted.

In any case when it comes to quantum pasts, as Duke Ellington taught us, “Things ain’t what they used to be.”
Perhaps Bernstein wrote this short paper just for that final joke at the end?

To summarise, if you take quantum theory seriously and you take eternalism (the block universe) seriously, then it seems that the past is as indeterminate as the future.

---

How does this relate to the 'low entropy in the past' idea used to explain the 'arrow of time'?

Since quantum theory is consistent with the second law of thermodynamics, a picture emerges of backwards branching towards (superpositions of) Big Bang variants.

My ex-colleague Roy said as much in this comment, on an earlier post devoted to the MWI. See here for more.

No comments:

Post a Comment

Comments are moderated. Keep it polite and no gratuitous links to your business website - we're not a billboard here.