This may be more of a philosophical question than a physics question, but here goes. The standard line is that nonrenormalizable QFT's aren't predictive because you need to specify an infinite number of couplings/counterterms. But strictly speaking, this is only true if you want your theory to be predictive at all energy scales. As long as you only consider processes below certain energy scales, it's fine to truncate your Lagrangian after a finite number of interaction terms (or stop your Feynman expansion at some finite skeleton vertex order) and treat your theory as an effective theory. Indeed, our two most precise theories of physics - general relativity and the Standard Model - are essentially effective theories that only work well in certain regimes (although not quite in the technical sense described above).

As physicists, we're philosophically predisposed to believe that there is a single fundamental theory, that requires a finite amount of information to fully specify, which describes processes at all energy scales. But one could imagine the possibility that quantum gravity is simply described by a QFT with an infinite number of counterterms, and the higher-energy the process you want to consider, the more counterterms you need to include. If this were the case, then no one would ever be able to confidently predict the result of an experiment at arbitrarily high energy. But the theory would still be completely predictive below certain energy scales - if you wanted to study the physics at a given scale, you'd just need to experimentally measure the value of the relevant counterterms once, and then you'd always be able to predict the physics at that scale and below. So we'd be able to predict that physics at arbitrarily high energies that we would have experimental access to, regardless of how technologically advanced our experiments were at the time.

Such a scenario would admittedly be highly unsatisfying from a philosophical perspective, but is there any physical argument against it?

  • $\begingroup$ Closely related: physics.stackexchange.com/q/295346/50583 $\endgroup$ – ACuriousMind Dec 20 '16 at 0:26
  • 1
    $\begingroup$ @ACuriousMind I think my question is getting at something at something different from that one. I understand how effective theories work when you have an energy cutoff above which new physics emerges. I'm asking whether it's possible that no (qualitatively) new physics ever emerges above your cutoff, and so your infinite series of nested effective theories is, in fact, the final theory. $\endgroup$ – tparker Dec 20 '16 at 0:44
  • $\begingroup$ What you say sounds an awful lot like how we can use Newtonian physics in cases where we don't have access to a fast enough (or small enough) test object to realize relativistic or quantum effects. $\endgroup$ – Cort Ammon Dec 20 '16 at 1:34
  • $\begingroup$ My apologies: I stole a line from your post above for my much more naive question : physics.stackexchange.com/q/299729 I hope you don't mind, it's much better expressed that I could manage. $\endgroup$ – user139561 Dec 20 '16 at 1:47
  • $\begingroup$ Question: by a nonrenormalizable QFT you mean a model defined by the Lagrangian on the background Minkowski spacetime? Minkowski spacetime is a solution of classical Einstein equations and is probably drastically modified by a quantum gravity theory. Quantum gravity spacetime is expected to be highly fluctuating on the microscopic scale. $\endgroup$ – Prof. Legolasov Dec 20 '16 at 6:36

You suggest that we can use a nonrenormalizible theory (NR) at energies greater than the cutoff, by meausuring sufficiently many coefficients at any energy.

However, a general expansion of an amplitude for a NR that breaks down at a scale $M$ reads $$ A(E) = A^0(E) \sum c_n \left (\frac{E}{M}\right)^n $$ I assumed that the amplitude was characterized by a single energy scale $E $. Thus at any energy $E\ge M$, we cannot calculate amplitudes from a finite subset of the unknown coefficients.

On the other hand, we could have an infinite stack of (NR) effective theories (EFTs). The new fields introduced in each EFT could successively raise the cutoff. In practice, however, this is nothing other than discovering new physics at higher energies and describing it with QFT. That's what we've been doing at colliders for decades.

  • $\begingroup$ I am not suggesting "that we can use a nonrenormalizible theory (NR) at energies greater than the cutoff, by meausuring sufficiently many coefficients at any energy." I emphasize in my question that we would need our experiments to go to higher and higher energies to measure the values of higher and higher counterterms. $\endgroup$ – tparker Dec 20 '16 at 23:09
  • $\begingroup$ The "infinite stack of EFT"s idea is what I am suggesting. But the point is that I am proposing that the higher-energy EFT's would not contain any "new fields" or qualitatively "new physics" - just more coupling constants for higher-order graviton interactions. Under my proposal, no new particles (involving gravity) would ever be discovered. $\endgroup$ – tparker Dec 20 '16 at 23:10
  • $\begingroup$ I don't understand what you mean. As you approach $E\lesssim M$, you need to know/measure more and more coefficients to make predictions. But once you surpass $E \ge M$, you need infinitely many, as all terms in the perturbation expansion are big (and in fact perturbation theory breaks). $\endgroup$ – innisfree Dec 20 '16 at 23:26
  • 1
    $\begingroup$ In my infinite stack of EFTs, massive fields are integrated out at each threshold, and each EFT includes all terms (R and NR) consistent with symmetries. Thus two EFTs with the same field content and symmetries are identical, and cannot differ by e.g. 'more coupling constants for higher-order graviton interactions'. $\endgroup$ – innisfree Dec 20 '16 at 23:30
  • 3
    $\begingroup$ That's the problem - above the cutoff of the EFT, all the high-order NR interactions become important at once. It isn't the case that you raise $E\gtrsim M$ and one NR interaction is important, raise it a bit more 2 NR interactions are important etc. $\endgroup$ – innisfree Dec 20 '16 at 23:32

The problem with GR+QM is that the counterterms include higher derivatives terms, $$ \mathcal L_\mathrm{ctr}\sim \partial^4h $$ where $g_{\mu\nu}=\eta_{\mu\nu}+h_{\mu\nu}$.

Therefore, on accounts of the Ostrogradsky instability theorem, the system is unstable. This means that the whole program of perturbation theory makes little sense, and there is no reason for us to expect that the perturbative expansion has anything to do with what the theory is really telling us.

Therefore, QGR's perturbative expansion need not reflect what the non-perturbative theory is about. We just don't know what to do with the theory, we only know that perturbation theory cannot work.

For more information on Ostrogradsky's theorem and its relation to a possible theory of quantum gravity, see the very accessible review Avoiding Dark Energy with 1/R Modifications of Gravity, by R. P. Woodard.

To close this answer I would like to include a transcript from Strominger's essay Is there a Quantum Theory of Gravity:

The problem of unifying quantum field theory and general relativity was formally solved in 1967 by DeWitt in a series of classical papers [...]. Using standard principles of quantum mechanics, he constructed a unitary set of Feynman rules describing the quantum dynamics of the gravitational field. Unfortunately, the perturbation series in Newton's constant is not renormalizable. This is evident from power counting: the naive degree of divergence of an $L$ loop diagram is $2(L+1)$. Barring miracles, we therefore expect on dimensional grounds that a counterterm constructed from $L$ powers of the Riemann tensor will be necessary to renormalize the $(L+1)$th order of perturbation theory. Since a new coupling constant is introduced at each order, the theory loses its predictive power. This difficulty arises from the fact that the coupling constant has dimensions of inverse mass.

A first response to this problem was to conjecture that the full theory is in fact finite [...]. After all, the perturbation expansion is really an expansion in the dimensionless parameter $\kappa^2E$ ($E$ is an energy scale, $\kappa^2=32\pi G$). This expansion parameter is large at large energies. One cannot expect such an expansion to provide a systematic method for computing the effects of virtual gravitons of arbitrarily large energies. The non-renormalizability of the weak coupling expansion may just be due to a bad expansion of a good theory.

The finiteness conjecture has both intuitive and calculational motivations. On the intuitive side, there is a vague notion that one should not be able to propagate at energies where wavelengths are less that the Schwarzschild radius. The Planck energy should thus provide a natural cutoff. This notion has also received some support from explicit calculations. Non-perturbative summations of ladder [...] and cocoon graphs [...] have in fact produced finite results.

Unfortunately, these calculations are not gauge invariant and do not represent a systematic expansion in some small parameter. Furthermore, in recent years, systematic expansions have been developed in parameters that are small at all energies [...]. Quantum gravity is not finite order by order in these expansions.

The prospects for a sensible quantization of the Einstein action thus do not appear terribly bright. The next logical step is to alter the action by adding higher derivative terms such as $R^2$ [...]. In any case, one is forced to add these terms for renormalization.

This step is taken rather hesitantly because actions with four time derivatives generally describe theories that appear pathological even at the classical level. In fact, it can be shown that classical higher derivative gravity theories either have tachyons or negative energies for small, long wavelength fluctuations [...]. Thus they have pathologies that are evident on macroscopic length scales. How, then, can the quantum version of these theories possibly be reasonable? The miracle is that some, if not all, of these instabilities can be systematically eliminated from the quantum theory. This will be discussed further in the next section.

I really encourage the interested reader to have a look at the rest of the essay. In a nutshell, one can in principle formulate consistent theories of quantum gravity, but at a very high price. In fact, most people haven't heard about these formulations, which is a strong hint that the community doesn't consider them to be relevant nor useful solutions to the GR+QM problem. For one thing, they generate more questions than actual answers. But'll I leave the reader reflect about the situation themselves.

  • $\begingroup$ no reason for us to expect that the perturbative expansion has nothing to do with... i.e. there's a reason to expect that it has something to do with it? Double negation is confusing, especially in English... $\endgroup$ – Ruslan Dec 20 '16 at 10:11
  • $\begingroup$ Interesting! But I've heard many people emphasize that the naive quantization of GR works fine as an effective theory at low energies (i.e. far below the Planck energy $G^{-1/2}$) - how do you reconcile this with your result? $\endgroup$ – tparker Dec 20 '16 at 23:01
  • $\begingroup$ @tparker it very much depends on what you mean by works. I really recommend you to read How Far Are We from the Quantum Theory of Gravity? for an overview of the problems of QG. It is very pedagogic and yet insightful. $\endgroup$ – AccidentalFourierTransform Dec 20 '16 at 23:30
  • $\begingroup$ Good point. I often hear claims of quantum gravity being plagued with infinities and doomed, but this is only true in the context of perturbation theory. $\endgroup$ – Prof. Legolasov Dec 27 '16 at 3:33

I think the reason people don't like this idea is that to keep the same physics at lower energy scales the coefficients of the nonrenormalizable terms should grow as you define the theory at higher and higher energy scales $\Lambda$. This is the flip side of being an 'irrelevant' interaction. They might become infinite at a finite value of $\Lambda$, in other words a Landau pole. In that case you could not define the theory at higher energy scales while keeping the same physics you already have at lower energy scales.

If they do not become infinite but in fact converge to a finite value, an ultraviolet fixed point, then the theory really is well defined. This is actually a serious proposal for quantum gravity called asymptotic safety.

Basically you can keep pushing the effective theory and either you start to see new physics, or it works all the way and you have asymptotic safety. The option where it really does break down at some finite $\Lambda$ would lead to people studying the regulator. Is it a lattice theory or what? That would really be new physics too.

  • $\begingroup$ I agree that the existence of a Landau pole at finite energy would invalidate my proposal. And asymptotic safety would make things rather straightforward. But isn't there also a third option, where the coupling constants flow to unboundedly large values at high energy, but don't ever diverge at any finite energy scale? In that case, the theory would also be well-defined, although of course you couldn't use perturbation theory to study it at high energy. $\endgroup$ – tparker Dec 20 '16 at 22:51

If one interprets "physically" like positivists, there can be no physical argument against it - if the theory predicts the observable facts, everything is fine, and the number of epicycles does not matter. A preference for a theory with less parameters goes already beyond positivism, it has to rely on Popperian (philosophical) empirical content (predictive power).

In real physics, this is what the experimenters have to do - observe more "epicycles", more of the lowest order terms. And it is the job of the theoreticians to try to find a new theory which allows to get rid of all the epicycles already found.

What one has to expect is that trying to find new epicycles will fail beyond a critical length. Think about some lattice regularization as a typical theory where the continous approximation, similar to atomic theory, fails below the critical distance. If what you can "see" is, say, 1000 times the critical length, anything will look yet smooth and you will succeed with a few lowest order terms of an effective theory. At 100 times you will already need much more terms, at 10 times it will start to fail completely, and at the critical length itself even the lowest 10000 terms will not save the game.

So, roughly, if you need one more term to explain the observations, expect that you are already near the critical length where everything will fail, and you need a new theory.

In some sense, gravity is simply that first nontrivial (non-renormalizable) lowest order term, and the Planck length is the corresponding prediction where we see that it cannot be ignored anymore in QFT computations. And where we have, therefore, to expect that the theory starts to fail completely too.


If you had two different equations for small scales and large scales, then the obvious problem is the equation for large scales would be inaccurate for small scales and the equation for small scales would be inaccurate for large scales. But what's not obvious at first is that there is a middle ground between the large scale and the small scale where neither equation is correct. That middle ground inaccuracy affects the accuracy of supercomputer simulations of real world processes.


protected by Qmechanic Dec 20 '16 at 9:22

Thank you for your interest in this question. Because it has attracted low-quality or spam answers that had to be removed, posting an answer now requires 10 reputation on this site (the association bonus does not count).

Would you like to answer one of these unanswered questions instead?

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