What is the evidence against the triviality of pure QED and pure $\phi^4$?

Question

This question sprouted from Arnold's comments of this post. I've seen more theoretical evidences hinting at the triviality of pure QED and pure $\phi^4$ in 4D spacetime than the opposite, I'll be more explicit:

(1)Perturbative expansions to a few loops order, for both QED and $\phi^4$. But I confess I myself have only verified the 1-loop results.

(2)Claims from lattice simulations. I never verified this for myself, and lattice $\phi^4$ numerics seem to be well cited, while the lattice QED numerics less so.

(3)Nonperturbative constructive results on $\phi^4$, a rigorous proof of triviality is said to be only a few lemmas away from completion (I remember this claim from an online lecture of Arthur Jaffe).

Now by "pure" I mean constructions that only make use of the infrared terms, for example if I remember correctly, evidence (3) can't exclude---even with the completion of the unproved lemmas---the possibility for the existence of a scalar field theory with additional $\phi^6, \phi^8, \phi \Box^2 \phi$ terms etc, such that it flows to a pure $\phi^4$ in the infrared.

What are the supporting evidences for the opposite claims?

Comments

I'll collect my evidence; it may take a while...

---

@ArnoldNeumaier, sure take your time, it's bedtime here in my time zone...

---

I just noticed that some of what I know on this query is already in the thread ''What is the status of the existence problem for scalar QFT and QED? ''. However, in the mean time I learnt something more, so the new answers are still useful.

---

I am sorry to be ignorant, but what do you mean by "triviality of a theory"? Reducing the exact solutions to free ones or what?

---

@ColdCooler, it's just another way of saying Landau pole, except the perspective is a bit different by reversing the direction of renormalization group flow.

---

The difference $f(\infty)-\sum_{n=0}^{\infty}f^{(n)}x^n/n!, \; x\to\infty$ is exactly zero, but its "perturbation series" is non trivial and divergent. The Taylor series, if summed up selectively (not all its terms, but "the most divergent" ones), may result in something different from zero. That may be the origin for Landau-pole behavior of such "exact" sums.

Answer 1

The most standard argument for triviality is the Landau pole of QED and $\phi^4$ theory. This was a valid argument in the early dayys of renormalized QFT but it is no longer valid today. After all, QCD also has a Landau pole, and unlike QED, its Landau pole is at physically realizable energies. See, e.g., p.27 of https://arxiv.org/pdf/hep-ph/9802214.pdf. But nobody ever suggested that this is a reason for the triviality of QCD. Similarly, quantum Yang-Mills theory has a Landau pole, https://arxiv.org/pdf/1311.6116.pdf, but it is believed that the theory exists rigorously, and a proof of it would solve half of the 7th Clay Millennium Problem (the second half being a proof of a mass gap for it).

The Landau pole is most likely an artifact of perturbation theory, caused by the latter's lack of accounting for terms with an analytic dependence of the form $e^{const/\alpha}$, where (for QED) $\alpha$ is the fine structure constant. Evidence for this is the fact that the Landau pole in 1-loop perturbation theory for QED is cancelled exactly by a second, nonperturbative term that does not contribute perturbatively. As a result, the spectral function is well-defined and satisfies the causality conditions required by the Källen-Lehmann representation. See the paper by Bogolyubov, Logunov and Shirkov, The method of dispersion relations and perturbation theory, Soviet Physics JETP 37 (1960), 574-581. There is a highly cited review paper of the resulting analytic perturbation theory (concentrating on QCD, though) by Shirkov and Solovtsov, Ten years of the analytic perturbation theory in QCD. Theoretical and Mathematical Physics 150 (2007), 132-152. An interesting recent paper on the subject (for QCD) is https://arxiv.org/abs/1411.2554.

A 2017 paper by Djukanovic et al. shows that treating QED as an effective theory by adding an additional Pauli interaction term already removes the Landau pole at 1-loop order. 

For $\Phi^4$ theory, my positive evidence is somewhat limited, since this is somewhat aside from my main goal, to prove the existence of QED. 

My main argument for $\Phi^4$ theory is Section 8, ''IS DESTRUCTIVE FIELD THEORY POSSIBLE?'' in a paper by Gallavotti and Rivasseau from 1984, which discusses $\Phi^4$ theory in 4 dimensions and is very pessimistic that the existence of a rigorous $\Phi^4_4$ quantum field theory can be ruled out. The main reason given is that one needs arguments valid for all possible constructions, while the usual approach (leading to negative evidence) is to consider the simplest classes of approaches only. As far as I can tell no work at all has been done since then on their Super-strong Triviality Conjecture, for which they state ''we do not see at the moment any compelling reason to believe it at all''. (This is confirmed by a remark of Gallavoti at the end of p.14 in a 2014 paper:

the conjecture that it is impossible to obtain nontrivial Schwinger functions in a scalar quantum field theory in dimension 4 is still (wide) open

Callaway's 1988 comprehensive review article on triviality comments the first of these papers on p.290 as follows: 

Other attempts to construct a nontrivial $\phi^4$ theory are often a bit more abstract in nature. Rigorous discussions of triviality (see section 2) often require that a $\phi^4$ field theory is defined as an infinite-cutoff limit of a ferromagnetic lattice theory. It has been argued [4.31] that this is an assumption whose removal changes the nature of the problem dramatically. Indeed, no argument appears to prevent the existence of an interesting nontrivial ultraviolet limit of an antiferromagnetic lattice $\phi^4$ theory, even in d >4. This remains an interesting open problem.

While this is no direct positive evidence it is doubly negated evidence, which may count as positive, too. Klauder claims that an approach based on affine coherent states should work in any dimension, but his arguments are not rigorous.

See also my answer in the thread ''What is the status of the existence problem for scalar QFT and QED?'', which contains additional information.

Comments

@JiaYiyang:

whatever the final fundamental theory is, it must somehow approach something that contains the interacting QED at low energies

Whatever the final fundamental theory is (if one exists at all, which is by no means certain), it contains an effective vacuum state and effective fields describing the electromagnetic field $F_{\mu\nu}$ and the electron field $\Psi$, and a Poincare group acting on it. These fields generate a field algebra whose gauge invariant part satisfied the Wightman axioms and hence provides a model for QED at all energies, which coincides with standard renormalized QED at low energies. Possibly the standard renormalized QED field equations get (formally nonrenormalizable) higher order correction terms that can be ignored at low energies but not at higher ones. It is this version of QED that exists beyond reasonable doubt, due to experimental evidence and very general symmetry considerations. (Note that in 2D there are formally nonrenormalizable theories that nevertheless exist rigorously.)

However, beyond that, I think it is likely that this version of QED actually satisfies the renormalized QED field equations exactly.

---

Possibly the standard renormalized QED field equations get (formally nonrenormalizable) higher order correction terms that can be ignored at low energies but not at higher ones. It is this version of QED that exists beyond reasonable doubt

This I would agree, but this is a rather broad stroke of brush on the triviality problem. It is basically allowing for "new physics" at high energy scale to fix the triviality issue(the same applies to some of the citations you give). However, if we look along this direction, possibilities run wild, and the physics becomes a more urgent issue than the mathematics. It's quite possible that by adding enough delicately chosen terms/regularization methods to the QED Lagrangian, the Laudau poles disappear. But if the new Lagrangian is too complicated, then the question is "why bother?" Why bother putting all that effort just to avoid triviality? Why let the avoidance of triviality be a guiding principle for new physics? Very likely such effort gives a nontrivial theory, valid at all energy scales yet fail to describe the real universe. Isn't effort better spent on more pressing and rewarding issues like quantum gravity? 

On the other hand if the new Lagrangian is still reasonably simple, then I retain much mathematical interest on the issue, but it seems not to be the case?     

---

@Dilaton, but here we are not taking EFT point of view.

---

@Dilaton, unless you take the philosophical stand point of "There's no theory of everything at all", or you must have missed the entire point Arnold and I have been discussing. We are discussing the possibilities of the fundamental theory being a QFT,  while what you have been saying are correct but irrelevant stuff. 

---

@JiaYiyang: Referring to a fundamental theory only guarantees existence of QED possibly with infinitely many higher order terms. However, these additional terms then encode (part of) all other content of the fundamental theory. On the other hand, one would expect that any consistent theory containing QED would give one of these variants of QED, so that there should be infinitely many of them (unless there is only a unique theoretical possibility for the fundamental theory, which I doubt very much). This means that if one wants to construct QED without the fundamental theory one should have much freedom in choosing the higher order terms. Indeed, standard renormalization theory suggests that one can choose freely infinitely many parameters. But one needs to construct only one of these to have a good theory.

On the other hand, power counting arguments say that in perturbation theory, one gets a valid simplified theory form any theory with infinitely many terms by simply deleting all nonrenormalizable terms. I expect this statement to be valid rigorously. Thus if one of the QED variants incorporating some physics beyond QED exists, I expect that standard QED also exists.

As I had already mentioned, in causal perturbation theory a Landau pole doesn't force nonexistence. The Landau pole only forbids constructions that are based on an energy cutoff that must be moved to infinity (or a short distance cutoff that must be moved to zero) - beyond the pole. This forbids lattice approximations. Indeed the only positive triviality results are for (particular classes of) the latter only. But all other approximations that keep from the start the correct short-distance structure are unaffected by Landau pole arguments.

I do not really know what (if anything) to expect as consequences of a Landau pole in causal perturbation theory - One day I'll have to do corresponding calculations.

Isn't effort better spent on more pressing and rewarding issues like quantum gravity? 

I believe that the issues with quantum gravity will sort out themselves once the existence of QED is positively settled, and not before. I have many indications that suggest that the main problems to be overcome in both cases lie in the inadequateness of traditional renormalized perturbation theory rather than in the nonexistence of the canonical quantizations of electrodynamics or gravity.

The same techniques that will construct QED will also (though probably with considerably more work) construct the standard model, quantum gravity, and their unification.

This is why I think constructing QED is by far the most important open problem in theoretical physics. It surely is the oldest one, hence it has for mathematical physics the same status that the Riemann hypothesis has for mathematics. And it appears to me tractable enough that I study in detail all techniques that promise some contribution to this problem.

---

The exact statement is: If QED exists at low energies (which you accepted), it must exist at all energies. For existence means that the uncharged sector satisfies the Wightman axioms. From the Wightman axioms together with the massless photon field one can deduce that the spectrum of the momentum vector (generator of Poincare translations) is unbounded, and all (smeared) creation operators for any timelike momentum $p$ exist.

Wilsonian type renormalization group arguments do not apply to QED in causal perturbation theory, since there is no cutoff to be taken to infinity. A Landau pole can therefore at worst mean that there is no scaling limit, i.e., no limit of QED where the electron mass vanishes.

I'll review other positive evidence after returning form my holiday.

---

@ArnoldNeumaier, ok thanks, and enjoy the holiday!

Answer 2

Let me put my two pennies to this question about triviality and non triviality. I will reason as physicist, not as a mathematician.

Fist of all, we must understand what kind of exact solution we expect to obtain from a theory. QED, as well as $\phi^4$, are "self-interacting theories". It means we inevitably obtain some corrections to a free solution. What physically such corrections mean, what wrong with a free (a plain wave) solution is, that's the question.

You may have missed this fact, but the usual Classical Electrodynamics of point charges is "trivial" too: the self action leads to self-induction by the charge near field that resists to any changes in the charge motion in an external field. Mathematically, the self-induction results into an infinite addendum to the original (free motion) mass. The right way to handle this situation is to put the charge (the interaction) to zero, at least at this self-induction term because we did not want any self-induction in our project. We wanted a weak radiation reaction influence and for that we must try other, more physical, approaches. Isn't is evident?

If you count renormalization will handle this situation, then good luck to you because nothing is promised with such an approach. In other words, the desired physics is not implimented from the beginning.

Comments

"If you count renormalization will handle this situation, then good luck to you because nothing is promised with such an approach."

The very accurate quantitative success of renormalized QED shows that you are mistaken.

---

Arnold, have you ever seen a fluke in Physics? There are many.

---

I have never seen a fluke that was even close to being as successful in predictivity as QED. If you call QED a fluke then you speak a language different from anybody else.

---

Success of QED is not based solely on renormalization, and I have already expressed my vision of it many times. I call "success" of renormalization a flike, following P. Dirac. And please, do not exagerate QED "precision" because it really depends on what process we dscribe.

---

Success of QED is not based solely on renormalization, but apart fron the defining interaction, renormalization is its essential ingredient: If you take it away, only a mess of infinities remains.

I know your vision - it is a fluke that hasn't produced anything comparable to QED.

---

The very long history of QED development shows how difficult and weird it was to get something reasonable.

Renormalization is nothing but changing the calculation result. It is much better to have a theory that does not need modifying its results.

Returning to the original post question: it depends whether you accept, for example, $\phi^4$ as it is (then it is an ill theory with no results) or you are ready to modify its results until you get something that you like.

---

''Renormalization is nothing but changing the calculation result. It is much better to have a theory that does not need modifying its results.''

Causal perturbation theory has precisely what you desire. Nowhere any result is changed; everything is derived with full mathematical rigor (though perturbatively only) from a set of axioms. 

---

Arnold, we have already discussed it before, and I do not buy this "causal" approach - simply because the differential equations and the boundary cinditions are sufficient for causality, in my opinion. Scharf predends it is not so; thus he "works out" the solutions.

---

If you don't like Scharf then look at QED as done in the recent book

• Michael Dütsch, From Classical Field Theory to Perturbative Quantum Field Theory, Birkhäuser 2019

I think Dirac would be fully satisfied with Dütsch's account, since nothing mathematically surprising or offending happens anywhere.

Then tell me where he changes calculational results, or modifies any results, or whatever you didn't like about Scharf's treatment. Be specific about your criticism. In particular, please tell us what is wrong with "working out" solutions - a process necessary in all mathematical problems.

---

I do not have this book; I just browsed some of its pages and read its annotation available online. The author lists different pretexts for modifying the perturbative solutions proposed by different people at different times. Then he focuses on another pretext - keeping the maximum similarity to the Classical Field Theory, about which I expressed my skepsis and got a "-1" score for calling CED "trivial".

I do not have much health to work in this unhealthy atmosphere, so I quit this discouraging talk.

---

''we did not want any self-induction in our project. We wanted a weak radiation reaction influence and for that we must try other, more physical, approaches. Isn't is evident?''

No. ''We wanted'' is just you. Others Including Paul Dirac) just want to have a perturbative QED that coincides in its results with the traditional, highly successful QED and has no mathematical defects.

---

Arnold, QED is built by analogy with CED, in particular, the interaction term is the same in form $H_{\text{int}}\propto A\cdot j$, but with operators instead of functions. I just underlined that CED is not perfect either, so hoping that quantization may give something reasonable immediately is too brave, but unsubstantiated, hope. And indeed, apart from divergent mass corrections, in QED we obtained divergent charge corrections and the infrared catastrophe; everything needed some special simplifications and some particular "treatment" to obtain something to compare with experiment.

---

Nothing depends on the ill-defined connection between classical and quantum electrodynamics. The quantum case stands for itself., and avoids the problems of the classical theory.

Note that QED is not a quantization of CED - the field equations involve an interacting Dirac equation which does not figure in CED.

The IR problems of QED that you talk about disappear when one uses physical coherent states instead of unphysical free electrons as incoming and outgoing states. 

---

Nothing depends on the ill-defined connection between classical and quantum electrodynamics. The quantum case stands for itself., and avoids the problems of the classical theory.

QED at low energies must reduce itself to CED, and this reduction is discussed in many textbooks.

Yes, we need use "dressed" states, with more degrees of freedom rather than free particles, and as you know I proposed a toy model of such a dressed electron ("electronium"). Coherent states are somewhat inexact since they include states with infinite energies. My "dressed electron" may not get an infinite final energy after scattering from it a projectile with a finite initial energy.

---

It is quite common that approximations (such as CED as an approximation to QED suffer from problems outside their domain of applicability. Thus the difficulties of CED tell nothing about the problems of the QED it approximates in the low energy limit.

---

Thus the difficulties of CED tell nothing about the problems of the QED it approximates in the low energy limit.

It is a funny opinion. The lower energy, the more exact is approximation.You may not ignore it completely.

---

This is not true. A quantum theory does not approach a classical theory at low energy, not even an anharmonic oscillator does. Instead, the approximation gets better and better the bigger the masses involved.

Therefore CED is reliable at macroscopic masses, whereas the problems at the tiny electron mass are due to it being a poor approximation of quantum theory at this scale!

---

A quantum theory does not approach a classical theory at low energy... Instead, the approximation gets better and better the bigger the masses involved.

Bigger masses with respect to what? Low energy means low with respect to $mc^2$.

To be more specific, I mean scattering a point charge in an external field accompanied with radiation, and radiation reaction problem.

---

Bigger masses are masses of mesoscopic or macroscopic matter. Compared to that, an electron has a very tiny mass, very far from making it behave classically.
 

---

There are many electronic devices whose operation is described with CED.

And there was a huge scientifical challenge to solve the problem of radiation reaction in CED starting from H. Lorentz and onwards. You may not ignore these efforts of many.

---

But electronic devices whose operation is described with CED are based on macroscopic currents, not on that of a single pointlike electrons.

The problem of radiation reaction occurs only in CED, not in QED. Thus it can be ignored as a problem produced by approximating QED by CED in a domain where the latter is clearly wrong.

---

From your statement one may have an impression that QED resolves the problem, but it is not so.

---

From your statement one may have an impression that QED does not solve the problem. Where then are these efforts of many that address this huge scientifical challenge to solve the problem of radiation reaction in QED?

---

It first appeared as the IR catharstrophe; then one started to use Bloch&Nordsieck approximation (no recoil is taken into account). Since then this had no advances to take into account the soft radiation recoil :-(.

---

An infrared stable perturbative S-matrix for QED is constructed to 2 orders in perturbation theory in 

• Pawel Duch, Infrared problem in perturbative quantum field theory, https://arxiv.org/pdf/1912.04885.pdf

 There is not the slightest sign that soft radiation recoil would cause any problems. Such problems solely come from using unphysical asymptotic states when trying to evaluate S-matrix elements. 

---

These are two different links to two original papers. Correct it for them to be so.

These papers are of 2019 and 2020, so this question is still on the table.

What I see is not similar to my comprehenion of the QFT problems; I outlined my vision in my papers on arXiv. So far I haven't seen any feedback (except for this site negative scores), so, I guess, there will be still many lemmas in perspective within the wrong guess of QFT fundamental constituents.

But I do not care anymore. Too late, Arnold, too late. Time is not reversible, and I have no desire to expose myself to your harsh, but wrong, evaluations.

Take care.

Bob.