I am most interested in interpretational differences due to the fact that quantum field theory is relativistic while quantum mechanics is not. By "Copenhagen interpretation" I mean a minimal interpretation that connects mathematical formalism to observations.

The answer to Collapse in Quantum Field Theory? says:"The collapse of a wavefunction - or its decoherence, or splitting off into different branches as it gets entangled with a measurement apparatus - looks exactly the same". But this is odd. Quantum mechanics has a global time variable, so it makes sense to talk about quantum state at time $t$ being a superposition, and then being a collapsed eigenstate at a later time $t'$. Of course, it turns out that although QM is non-relativistic and collapse is formally "instantaneous", where physical entities are concerned it happens to be compatible with relativity by happy coincidence.

But in a relativistic QFT such description does not work even formally. There is no global time or absolute simultaneity, so no "quantum state at time $t$" that can be described as collapsing. One could try to relativize this to a particular observer, but such "relative collapses" are incoherent because different observers have different simultaneity surfaces. So in QFT instantaneous collapse would not just be formally non-relativistic, but meaningless like a syntax error. So how is the collapse (whether actual or apparent) interpreted in QFT in a way consistent with special relativity?

EDIT: After searching I found Reality, Measurement and Locality in Quantum Field Theory helpful, it analyzes the EPR experiment from the QFT point of view, and discusses collapse explicitly. On interpretational issues of QFT more broadly Against Field Interpretations of Quantum Field Theory gives a nice overview.

  • 1
    $\begingroup$ Did you have a look at the classic paper by Bohr and Rosenfeld? $\endgroup$ – Tom Heinzl Jul 14 '15 at 8:56
  • $\begingroup$ @Tom Heinzl Thank you, a very interesting paper. Unfortunately, BR seem to focus on aspects of measurement which are not in apparent conflict with relativity, like collapse. $\endgroup$ – Conifold Jul 16 '15 at 23:08
  • 1
    $\begingroup$ This paper may be of interest: journals.aps.org/prd/abstract/10.1103/PhysRevD.21.3316 $\endgroup$ – Nirmalya Kajuri Nov 13 '15 at 23:39
  • $\begingroup$ also answered here (but seems to differ with current answers) physics.stackexchange.com/questions/94385/… $\endgroup$ – user83548 Nov 17 '15 at 16:05
  • $\begingroup$ Note that the claim ''wavefunctional space is unitarily equivalent to many-particle Fock space'' made in the abstract of the paper ''Against Field Interpretations of Quantum Field Theory'' referenced in your question is invalid for interacting theories. It is true only with a - nonphysical - cutoff. But this cutoff is removed by renormalization, where the unitary transformation becomes ill-defined. Indeed, renormalization shifts the unphysical infinities from the S-matrix elements to the unitary similarity transformation. $\endgroup$ – Arnold Neumaier Nov 17 '15 at 17:03

The Born rule (and hence any discussion of collapse in the sense of the Copenhagen interpretation) is relevant only when an observer has made a distinction between a (tiny, observed) system and its (huge, observing) environment (= everything else, containing in particular the measurement equipment).

This distinction (not present in relativistic QFT itself) already breaks Lorentz invariance. The collapse (describing conditional probabilities conditioned on observations) is a property not of the wave functional in QFT but of its restriction to the Hilbert space of the observed system, which is an observer-dependent, vanishingly small part of the Hilbert space of the complete (observed + measuring) system.

This restricted few particle system is only an effective theory, to which fundamental considerations cannot be applied.

Thus there is no contradiction. A sequence of papers with the title Classical interventions in quantum systems by Asher Peres discuss how observations by different observers remain consistent in a relativistic framework.

  • $\begingroup$ see also physics.stackexchange.com/a/219163/7924 for how to obtain the effective dynamics of the restricted subsystem in a particular case. $\endgroup$ – Arnold Neumaier Nov 18 '15 at 12:35
  • $\begingroup$ This seems like a suggestion that QFT solves the quantum measurement problem. Am I misinterpreting, or is that correct? $\endgroup$ – AGML Nov 18 '15 at 20:38
  • 1
    $\begingroup$ Thank you, Peres definitely goes to the heart of this. But he sets the bar for interpretation very low, all that has to be relativistic are probabilistic predictions. But QFT formalism is manifestly relativistic, so any interpretation will do that much. Quantum states and collapses in QM are unphysical, but at least invariant across observers, representing "maximum knowledge" under Copenhagen. Peres emphasizes that states and operators are not just non-physical, but have "no real meaning". If nothing invariant is added beyond the bare formalism doesn't it make such interpretation redundant? $\endgroup$ – Conifold Nov 19 '15 at 0:13
  • $\begingroup$ Statistical/Feynman interpretation, that Tommasini mentions, takes ensemble amplitudes as basic and explicitly eliminates all non-invariant elements from the picture. Attraction of Copenhagen for QM seems to be that it provides some "overall picture" behind probabilities, albeit deeply non-classical. If this does not carry over to QFT shouldn't we adopt Feynman? $\endgroup$ – Conifold Nov 19 '15 at 0:13
  • 1
    $\begingroup$ Invariant amplitudes only give the S-matrix, which is no longer dynamical. To have a good dynamics one would need a renormalized perturbation theory for arbitrary correlation functions, not only time-ordered ones. $\endgroup$ – Arnold Neumaier Nov 20 '15 at 6:25

I take a minimal interpretation of QFT in a Copenhagen style to seek to make a connection between a classical description of/model for an experimental apparatus and classical records of its measurement results and a QFT model for the same apparatus.

Classically, a modern measurement device is most often a thermodynamically metastable system that we engineer to make transitions from a Ready state to a Detected state, and for which we also engineer an explicit feedback that brings the state back to Ready as soon as possible. For such a device, electronics detect a change of voltage from 0V to 1V, say, and makes a classical record of the approximate time at which the transition happened (and, perhaps, of various classical settings of the apparatus at the time; see Weihs's Bell experiment for a concrete fairly straightforward example, http://arxiv.org/abs/quant-ph/9810080). Typically we make millions of such classical records, which we group together in some way or another to construct ensembles (for Weihs, two events at close enough to the same time = one element of the highest-level ensemble, which can be split into 16 sub-ensembles according to the recorded classical settings). From this, we can construct statistics and show that they correspond or do not correspond well to whatever QFT models we may have constructed for the experiment (for the simplest cases, QFT is pretty much just quantum optics, we don't have to worry much about the interacting QFT of the later-added part of your question, and the asymptotic fields associated with S-matrix results are about as straightforward).

There is a classical more-or-less continuous signal that underlies the discrete events, which hardware and software converts into times when thermodynamic transitions happened (for the sake of storage limitations because recording the signal picosecond-by-picosecond would be enormous and likely irrelevant). The signal is rather imprecise, in that it's not an observable quantum field along a time-like trajectory, which is not possible because of the field commutation relations, but is instead a functional of thermodynamically large numbers of DoFs associated with the measurement device, for which field commutation relations have far less effect than the change from 0V to 1V that signals a measurement event. Nonetheless, we take it that the statistics of events are coupled to the rest of the experimental apparatus and will be changed by any change to the rest of the experimental apparatus. Whatever changes there are to the recorded statistics can be modeled by choosing a different state of the quantum field (or alternatively by choosing a different operator). For a given measurement operator, we can perhaps reasonably say that the state of the quantum field "causes" the observed statistics to be what they are (rather close though this is to commonly discounted ensemble interpretations of QM, https://en.wikipedia.org/wiki/Ensemble_interpretation), but perhaps it's as well to be more reserved when choosing whether to claim that the quantum field causes individual observed events.

From this point of view, the "collapse" is a classical property of an experimental device that we have engineered to be in a thermodynamically metastable state. If one also takes the view that QFT is an effective field theory that is essentially stochastic, the Lorentzian dynamics are a property of the statistical-macroscopic level of model, so that we cannot make any direct claim about the symmetries of the dynamics at the level of individual events. Indeed, we know that the macroscopic effective dynamics of superfluid Helium, being Lorentzian but with the speed of sound replacing the speed of light, is significantly different from the microscopic dynamics, so we should not rush to assume that the dynamics associated with individual events has the same symmetries as the dynamics associated with the statistical level dynamics. This is not to claim that there is a particular FTL deterministic dynamics, potentially associated with a different metric as in the superfluid Helium case, but it leaves the door open for one, which is all I feel the need for because I'm mostly content just to use QFT; if you want a specific choice of dynamics at the level of individual events, that's harder. Current experiments are very far from ruling out all possible classical local dynamics, they can only rule out the straw man of Lorentzian dynamics.

Perhaps we can also reasonably note that modern Quantum Gravity approaches give up Lorentzian dynamics at Planck scales with the intention that we will be able to show that the effective dynamics at large scales will nonetheless be Lorentzian.

You will note that the above doesn't much engage with GRW as it is usually described, for which collapse is not nearly so tied to experimental details as I have it above, which I suggest is more as a Copenhagen-style interpretation should have it. The distinction between stochastic/statistic and deterministic levels of description is of course problematic in its evocation of Einstein's later worries about quantum theory, which I suggest, however, can be revisited with modern ideas about effective field theories in mind, if one cares enough and can think of a way to do it.

  • 2
    $\begingroup$ Thank you. The search for underlying deterministic or at least "real" dynamics was never sufficiently motivated for me, it seems like looking for a classical picture underneath simply because we are used to it macroscopically. But if even Lorentzian picture is only effective what reason is there to expect a classical one? So perhaps already QM Copenhagen admits too much with quantum states and operators, while QFT suggests that nothing beyond probabilities of events conditioned on other events has even theoretical significance? $\endgroup$ – Conifold Nov 19 '15 at 0:29
  • $\begingroup$ A classical PoV is just a different PoV; if we think in classical stochastic field terms as well as in the usual QFT terms we may notice things that we otherwise would not, but of course it might be a waste of time. We each just choose whatever approaches we individually think look hopeful. No reason to "expect" a classical underlying dynamics, but anyway we can pursue alternatives. The idea that the dynamics is Lorentzian is an empirical principle, so it might not be correct at all scales, indeed GR has it as not correct at large scales. $\endgroup$ – Peter Morgan Nov 19 '15 at 2:56
  • $\begingroup$ @Conifold, last sentence: I'm not quite sure what you're saying. I'd say QFT introduces states and operators as a mathematical model for statistics and correlations between events. As something of a riff on what I think you might mean, I find Lakatos's idea of "bridge principles", pragmatic connections between absolutely raw experimental data and theoretical models, is good enough for my relatively modest philosophical purposes. In such terms, "significance" and/or meaning of mathematical elements of a theory are not absolutely clear, but we just describe the connection as well as we can. $\endgroup$ – Peter Morgan Nov 19 '15 at 3:08
  • $\begingroup$ ''that the dynamics is Lorentzian is an empirical principle, so it might not be correct at all scales, indeed GR has it as not correct at large scales'' - This is not true; Lorentz symmetry is an exact gauge symmetry of general relativity, valid in each orthonormal frame. This is especially visible in the Palatini formalism needed for handling fermions. GR only deforms the the translations of the Poincare group (into arbitrary diffeomorphisms). $\endgroup$ – Arnold Neumaier Nov 19 '15 at 15:58
  • $\begingroup$ For example, image charges in electrostatics are mathematical fictions and play no role in interpreting how the field is formed. But that is not how Bohr treats QM wave functions, they are non-physical but not fictional, and interpreted as "maximal states of knowledge". A minimal condition for such significance seems to be that one can at least coherently define the concept in a theory, as physical or not, which is not the case in QFT due to frame issues. Unless one is also willing to adopt a privileged frame, a.k.a. "fixed background", but that would be counter to either Bohr or Einstein. $\endgroup$ – Conifold Nov 19 '15 at 18:31

Quantum theory has a preferred time, and QFT is nothing but a standard quantum theory. So it has a global time. Because of the violation of Bell's inequality, every realistic interpretation needs a global time. That the theory is relativistic has only the consequence that different versions, with different choices of the global time, do not lead to different physical predictions. But a realistic collapse interpretation would have to have a global time. And it has one.

A description based on a preferred frame does not have to care about what some observers think is their rest frame, but would be based on an objective rest frame. It would be unobservable. Which would be, for positivists, sufficient to reject its existence. But for realists, this does not matter, Nature is not obliged to make everything what really exists observable to humans.

  • $\begingroup$ I can't make much sense out of this answer. I can't tell if you disagree with Arnold Neumaier's accepted answer. If so, why? If not, then what does this answer add? $\endgroup$ – Ben Crowell Jul 7 '17 at 16:37
  • $\begingroup$ I have edited the comment, to make it more clear. Neumaier's answer imho does not address this error in the question that something which works in QT would not work in QFT. Neumaier's claim that some distinction between system and environment breaks Lorentz invariance is imho wrong, QT is inherently not Lorentz-invariant. $\endgroup$ – Schmelzer Jul 8 '17 at 20:08

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.