Does renormalization means discarding corrections to a known constant?

[Bob_for_short]

Dear experts,

Does renormalization mean discarding corrections to a known constant?

I mean, we assign a known value to the electron mass or charge, whatever, in the zeroth order of the perturbation theory, for example, in QED. In the next order we obtain a correction to this value (finite, for instance). AFAIK, the renormalization means replacing the known constant+correction by the same known constant. To justify this, one says that the constant is not known anymore but "bare" now and only the sum is observable.

In my opinion, this replacing may be understood without ambiguity as discarding the correction and keeping the known value intact. Is it so?

Thanks,

Bob.

 

[meopemuk]

Renormalization in QED is a bit more complicated than just manipulation with constants (electron's mass and charge).

When we calculate the QED S-matrix with the original Hamiltonian in high perturbation orders, we find that matrix elements (scattering amplitudes) come out infinite. The deep reason for this is that QED Hamiltonian contains self-interactions of particles. For example, a single electron self-interacts with itself, so that the scattering amplitude of the innocent process electron->electron is infinite already in the 2nd perturbation order.

The renormalization theory then says that the original QED Hamiltonian is bad, and it should be replaced by another Hamiltonian, which has additional "counterterms". The actual form of these counterterms is chosen by imposing two physically reasonable conditions. The first (mass renormalization) condition tells that the scattering amplitude of a single particle (e.g., electron) should be exactly the same as if this was was a free non-interacting particle. The corresponding counterterms then cancel all self-interaction effects in the S-matrix. The second (charge renormalization) condition tells that in the low-energy limit the scattering amplitude of two charged particles should tend to the value known from classical electrodynamics. For example, this amplitude should be proportional to the charge squared e^2, and should not have any contributions from higher orders, e.g. e^4.

It appears that there are three types of counterterms needed to satisfy these conditions. Counterterms of one type are designed to cancel infinite contributions from electron self-energy loops in Feynman diagrams. The other counterterm type cancels divergences due to photon self-energy loops. Counterterms of the third type cancel divergences in vertex loops. The counterterm interaction operators are formally infinite, so in order to perform calculations we need to introduce artificial cutoffs in momentum integrals. This is called regularization. In the end, the cutoffs should be removed, and, hopefully, all scattering amplitudes should remain finite in this limit.

After adding these counterterms to the QED Hamiltonian, one finds that, indeed, all divergences in the S-matrix get canceled by counterterms, and this cancelation is valid in all perturbation orders. However, counterterms do not cancel exactly the contributions from loop integrals. Some residual (finite) terms remain in each perturbation order. These terms are called "radiative corrections". They appear as remainders after subtractions infinity-infinity, and their calculation is the major task of the renormalization theory. The calculation of these radiative corrections in high perturbation orders is what made the renormalized QED the most accurate physical theory ever proposed.

As far as I know, there is no explanation why this messy renormalization process actually works. It does lead to very accurate predictions, but the reason for this success is rather mysterious. Another problem is that by adding counterterms to the Hamiltonian of QED, this operator becomes formally infinite (in the limit of removed cutoffs). This is a high price we pay for obtaining finite and accurate S-matrix. Without a well-defined finite Hamiltonian, renormalized QFT is not a complete and self-consistent theory. For example, in QFT it is impossible to evaluate the time evolution of states and observables. Bound states and their energies cannot be obtained (as in ordinary quantum mechanics) by diagonalization of the Hamiltonian.

 

[Bob_for_short]

Thank you, Eugene, but I know all that. That is why I proposed an explanation and a solution in my "Reformulation instead of Renormalizations".

My question is kind of Yes-or-No question. I would like to learn the answer.

Bob.

 

[DarMM]

Renormalization in QED is a bit more complicated than just manipulation with constants (electron's mass and charge).

Unless I'm mistaken, is this not exactly what renormalization is on the formal level?
Shifting the constants to have infinite values corresponding to divergences.

The first (mass renormalization) condition tells that the scattering amplitude of a single particle (e.g., electron) should be exactly the same as if this was was a free non-interacting particle.

Not to nit pick, but the condition is that it should have a pole at the physical mass. The two-point function will be very different from the free field theory. Unless I've misunderstood you and something else is meant.

As far as I know, there is no explanation why this messy renormalization process actually works. It does lead to very accurate predictions, but the reason for this success is rather mysterious. Another problem is that by adding counterterms to the Hamiltonian of QED, this operator becomes formally infinite (in the limit of removed cutoffs).

Actually it makes it rigorously finite in the limit of removed cut-offs. See the lovely paper of Glimm and Jaffe "Infinite Renormalization of the Hamiltonian is necessary".

 

[DarMM]

Thank you, Eugene, but I know all that. That is why I proposed an explanation and a solution in my "Reformulation instead of Renormalizations".

My question is kind of Yes-or-No question. I would like to learn the answer.

Bob.

I think perturbative renormalization, on the physical level, is the process of working out order by order what the bare mass should be in order that the physical mass is the value you know from experiment. It doesn't really involve ignoring things.

 

[Bob_for_short]

The question is about equivalence of renormalizations (whatever ideology is used) and discarding corrections. Does anybody (an expert) know the answer?

Bob.

 

[Bob_for_short]

I think perturbative renormalization, on the physical level, is the process of working out order by order what the bare mass should be in order that the physical mass is the value you know from experiment. It doesn't really involve ignoring things.

There is no bare mass in the zeroth order; only an observable mass m (it can be taken equal to 1). (The same for the electron charge.) There is a correction to a known fundamental constant in the next PT order. It is discarded and the result (a scattering amplitude, an energy level, whatever) becomes better than in the zeroth approximation. Is it so?

Bob.

 

[DarMM]

There is no bare mass in the zeroth order; only an observable mass m (it can be taken equal to 1). (The same for the electron charge.) There is a correction to a known fundamental constant in the next PT order. It is discarded and the result (a scattering amplitude, an energy level, whatever) becomes better than in the zeroth approximation. Is it so?

Bob.

No, I don't believe so. In the theory there is the bare mass, which must be a function of the cutoff in order that every correlation function is not identically zero. The problem is what function of the cutoff? Usually you use the renormalization conditions and perturbation theory to figure this out and it so happens that at zero order it equals the physical mass. At next order you see it's the physical mass plus some function of the physical mass and so on. However your not discarding anything, you're just working out the bare mass.

 

[meopemuk]

The first (mass renormalization) condition tells that the scattering amplitude of a single particle (e.g., electron) should be exactly the same as if this was was a free non-interacting particle.

Not to nit pick, but the condition is that it should have a pole at the physical mass. The two-point function will be very different from the free field theory. Unless I've misunderstood you and something else is meant.

You are right that usually one imposes two conditions on the electron propagator in the renormalized theory: The propagator must have a pole at the physical electron mass, and the residue must be equal to 1. However, propagators are not physical quantities. They are just abstract factors in Feynman integrals. So, I thought it would be more helpful to express the mass renormalization condition through something more directly related to the experiment. In any realistic theory of scattering, a free electron should propagate without (self-)interaction. This means that the matrix element of the S-operator between two 1-electron states should be <p|S |p'>= delta(p-p'). This simple condition is violated in the naive QED. Its validity is restored by adding renormalization counterterms to the interaction Hamiltonian.

Another problem is that by adding counterterms to the Hamiltonian of QED, this operator becomes formally infinite (in the limit of removed cutoffs).

Actually it makes it rigorously finite in the limit of removed cut-offs. See the lovely paper of Glimm and Jaffe "Infinite Renormalization of the Hamiltonian is necessary".

Thanks for the reference. I'll check it out. Perhaps you can explain briefly how the Hamiltonian of renormalized QFT can be finite? The counterterms are divergent, right? So, by adding the divergent counterms to the original Hamiltonian, we make the full Hamiltonian divergent too. Right? Thanks.

 

[meopemuk]

Thank you, Eugene, but I know all that. That is why I proposed an explanation and a solution in my "Reformulation instead of Renormalizations".

My question is kind of Yes-or-No question. I would like to learn the answer.

Bob.

Hi Bob,

Most certainly, I misundestood your question. Reading your post I've got an impression that you've simplified the renormalization procedure to trivial adjustment of parameters m and e in each perturbation order. I've attempted to tell you that actual renormalization steps are more complicated. But you already know that, so I'm not sure what the question was.

 

[Bob_for_short]

In the theory there is the bare mass, which must be a function of the cutoff in order that every correlation function is not identically zero. The problem is what function of the cutoff? Usually you use the renormalization conditions and perturbation theory to figure this out and it so happens that at zero order it equals the physical mass. At next order you see it's the physical mass plus some function of the physical mass and so on. However your not discarding anything, you're just working out the bare mass.

The very first time the people encountered the mass "corrections" in the classical electrodynamics (Lorentz, Abraham) of the self-acting electron. As I showed in "Reformulation instead of Renormalizations" (see Introduction, formula (I3)), the mass renormalization meant discarding the correction to the known value and postulating the resulting, new equation as a dynamical equation.

QED Hamiltonian uses the same self-interaction ansatz, so the same problem appear. In the Feynman-Schwinger QED the corrections to the fundamental constants were discarded too.
The notion of a bare mass (or a charge) is a way to make discarding look as a legitimate prescription. But we all agree that in the zeroth order (in the initial and final states in a scattering problem, for example) the values of the fundamental constants are the experimental ones. For the zeroth order solutions and for their Lagrangian many features are implemented to describe the experimental properties: Lorentz invariance, gauge invariance, classical limits, CPT, spin-statistics, propagator poles at physical parameters, etc.

Then, the theory is constructed so that in the next order some "corrections" to these fundamental constants appear (self-action effects). If these corrections were even small, the next order solutions would not agree with the experiments since the new mass and charge are different from the experimental values. So these corrections are discarded and the initial fundamental constants are kept intact. As soon as this discarding is obviously not legitimate, they invented the notion of "bare" constants that "absorb" the corrections. There are many posterior ways to "justify" the renormalizations - effective field approach, counter-terms, R-operation, etc. They are nothing but the ways to say that it is the mass which is wrong, not the corrections to it. (Poor guilty mass, it is wrong, but we know how to "doctor" it.)

I state that with the same success we can say that the fundamental constants are right but corrections to them are wrong, and this is the reason of their discarding. They are just not necessary. This prescription is absolutely equivalent to the renormalization results, isn't it?

Bob.

 

[DarMM]

The very first time the people encountered the mass "corrections" in the classical electrodynamics (Lorentz, Abraham) of the self-acting electron. As I showed in "Reformulation instead of Renormalizations" (see Introduction, formula (I3)), the mass renormalization meant discarding the correction to the known value and postulating the resulting new equation as a dynamical equation.

That is true, of classical electrodynamics.

QED Hamiltonian uses the same self-interaction ansatz, so the same problem appear. In the Feynman-Schwinger QED the corrections to the fundamental constants were discarded too.

The same problem does not occur, rather a different problem occurs. That of multiplication of the fields at the same point. A distribution-theoretic problem.

But we all agree that in the zeroth order (in the initial and final states in a scattering problem, for example) the values of the fundamental constants are the experimental ones.

That's only true for theories where the fields are directly related to the particles. It is not true in QCD for instance where the bare mass of the Dirac(quark) fields is at no order equal to the experimentally measured mass of any particle in the theory. For instance hadrons or mesons.

There are many aposteriory ways to "justify" the renormalizations - effective field approach, counter-terms, R-operation, etc. They are nothing but the ways to say that it is the mass which is wrong, not the corrections to it.

No, renormalization is basically looking at your lagrangian and seeing that it contains distributions being multiplied. From distribution theory you know that powers can be defined. It turns out that for qft the well-defined powers are just the old undefined powers with "infinite constants" in front. We call these the bare mass, e.t.c. due to the power they multiply. It's difficult to work out these powers so we do it order-by-order using certain conditions to help us. The physical mass is a parameter of the theory used as such a condition. However at no point is anything discarded.

 

[Bob_for_short]

However at no point is anything discarded.

I do not say that you discard something. I say I keep the fundamental constants intact and discard corrections to them. Is it equivalent to your results? Yes or no question.

Bob.

 

[DarMM]

I do not say that you discard something. I say I keep the fundamental constants intact and discard corrections to them. Is it equivalent to your results? Yes or no question.

Bob.

Well if the bare mass was kept the same as the physical mass and all expressions of it at higher powers of the coupling were discarded the theory would not exist, which is very different from the conventional approach where it does exist.

 

[DarMM]

You are right that usually one imposes two conditions on the electron propagator in the renormalized theory: The propagator must have a pole at the physical electron mass, and the residue must be equal to 1. However, propagators are not physical quantities. They are just abstract factors in Feynman integrals. So, I thought it would be more helpful to express the mass renormalization condition through something more directly related to the experiment. In any realistic theory of scattering, a free electron should propagate without (self-)interaction. This means that the matrix element of the S-operator between two 1-electron states should be <p|S |p'>= delta(p-p'). This simple condition is violated in the naive QED. Its validity is restored by adding renormalization counterterms to the interaction Hamiltonian.

Oh that's fine, didn't see what you meant at first.

Thanks for the reference. I'll check it out. Perhaps you can explain briefly how the Hamiltonian of renormalized QFT can be finite? The counterterms are divergent, right? So, by adding the divergent counterms to the original Hamiltonian, we make the full Hamiltonian divergent too. Right? Thanks.

It sounds strange alright. Basically the original Hamiltonian is totally divergent as an operator on Fock space. It's not densely-defined, self-adjoint or semi-bounded, this is because powers of the field are undefined being distributions. The renormalized Hamiltonian is not because the divergences of field powerings is canceled by the counterterm divergences or because the counterterms move you to a non-Fock space where the Hamiltonian is defined.
See page 4 of the thread "Unbounded operators in non-relativistic QM of one spin-0 particle" located https://www.physicsforums.com/showthread.php?t=304711&page=4". I've tried to give an explanation.

 

[Bob_for_short]

Well, if the bare mass was kept the same as the physical mass and all expressions of it at higher powers of the coupling were discarded, the theory would not exist...

Your statement is wrong. Apart from corrections to the fundamental constants, there are other corrections to the solutions. I discard only those that "correct" fundamental constants. So there is a non trivial reminder after such a discarding. I do not discard all contributions of higher orders, I do not stay within the zeroth approximation of the field equation solutions.

Let us make it evident. Let us take a QED Lagrangian without counter-terms. And let us put the mass m=1 and the charge e=1 in the zeroth order. That is our guess of the Lagrangian and that are our units.

In the next order I obtain a finite correction, say 3, to the mass: 1+3. I see that in a weighting experiment my electron weights now more than 1*g. I do not like it and I discard the addendum 3 from this expression. So my prescription is 1+3 ->1.

You say that in the expression 1+3 the value 1 is not 1 anymore but uncertain (?). To fix it, you require ?+3=1, that is the same result as my prescription. We obtain the same results for amplitudes and energies because now we returned to the right constant values and nothing else is involved in the renormalizations. In your approach your value (?) changes from order to order to be always =1. In my approach I eliminate unnecessary addenda to the right phenomenological constants. We obtain the same solutions (with 1 for mass, for example).

Of course, these complications are due to self-action ansatz. I managed to reformulate the theory in terms of "interaction" ansatz, but it is another story.

I would like to make it evident that the constant renormalizations can be understood as discarding extra addenda to the right fundamental constants in the frame of the original Lagrangian.

Bob.

 

[DarMM]

Your statement is wrong. Apart from corrections to the fundamental constants, there are other corrections to the solutions. I discard only those that "correct" fundamental constants.

It is those additional terms in the expression for the bare mass which allow the theory to exist. With out them the theory does not exist. This is rigorously proven to be true.

If what you are discussing is looking at the expansions for the correlation functions and discarding those infinities which you know to be related to the values of the constants, you will still not get the right results because certain finite parts in the final renormalized expression come about due to the specific nature of cancellations between that appropriate expression for the bare constants to that order and the correlation functions.

 

[Bob_for_short]

It is those additional terms in the expression for the bare mass which allow the theory to exist. With out them the theory does not exist. This is rigorously proven to be true.

If what you are discussing is looking at the expansions for the correlation functions and discarding those infinities which you know to be related to the values of the constants, you will still not get the right results because certain finite parts in the final renormalized expression come about due to the specific nature of cancellations between that appropriate expression for the bare constants to that order and the correlation functions.

So your answer is "no". But in my opinion the "specific nature of cancellations" in your and my approaches give absolutely equivalent "renormalized" constants. Then we both obtain the same finite (improved) solutions in each higher order.

Can anybody else express himself on this subject?

Bob.

 

[nrqed]

Dear experts,

Does renormalization mean discarding corrections to a known constant?

I mean, we assign a known value to the electron mass or charge, whatever, in the zeroth order of the perturbation theory, for example, in QED. In the next order we obtain a correction to this value (finite, for instance). AFAIK, the renormalization means replacing the known constant+correction by the same known constant. To justify this, one says that the constant is not known anymore but "bare" now and only the sum is observable.

In my opinion, this replacing may be understood without ambiguity as discarding the correction and keeping the known value intact. Is it so?

From an operational point of view, yes that's correct (as long as it is done very carefully).


From a conceptual point of view, what really happens is that there are parameters in the Lagrangian which are, a priori, completely arbitrary. The only way to find out what they are equal to is to do some computation and compare the result to some physical quantity. This must be repeated, ordere by order in the loop expansion. This is what renomalization accomplishes.

 

[humanino]

As far as I know, there is no explanation why this messy renormalization process actually works.

Does this take into account the work Connes, Kreimer and Marcolli on the Hopf algebra of the renormalization group ?

 

[genneth]

Does this take into account the work Connes, Kreimer and Marcolli on the Hopf algebra of the renormalization group ?

I've tried to understand their paper, but going on what I can understand at the moment, I don't think they've solved anything about renormalisation yet. They provide a way to systematically renormalise, provided you can sum the 1PI diagrams. Personally, I do not find that a helpful conclusion, since even summation of 1PI leads to UV divergences.

Incidentally, on the original discussion, I would like to point at Ken Wilson's Nobel speech as an excellent intuitive take on what renormalisation is doing.

 

[meopemuk]

Incidentally, on the original discussion, I would like to point at Ken Wilson's Nobel speech as an excellent intuitive take on what renormalisation is doing.

Do I understand it correctly that in the Wilsonian approach QED is not considered a self-consistent closed theory? One cannot take the limit of infinite momentum cutoff, because of "space-time granularity", non-point nature of particles (strings?) or some similar yet unknown short-distance behavior. This seems to be consistent with my statement that the Hamiltonian of renormalized QED becomes infinite in the infinite cutoff limit.

 

[Bob_for_short]

From an operational point of view, yes that's correct (as long as it is done very carefully).

Thank you, NRQED, for your positive answer. The correction discarding was practised and called “renormalizations on go”. It boils down to the requirement that the renormalization factors (like Z, Z_1, etc.) are equal to 1. It is briefly described in “Quantum Electrodynamics” by Berestetskii, Lifgarbagez, Pitaevskii. It has some practical conveniences.

Of course, such a discarding is not legitimate mathematically and much of effort was spent to “justify” it. We know how many different “concepts” have been proposed to “justify” the renormalizations. But the only working “justification” or the real reason of practising it is the good agreement with experiments.

From a conceptual point of view, what really happens is that there are parameters in the Lagrangian which are, a priori, completely arbitrary. The only way to find out what they are equal to is to do some computation and compare the result to some physical quantity. This must be repeated, order by order in the loop expansion. This is what renormalization accomplishes.

That is one of the “concepts”. It denies the original Lagrangian from the very beginning: they say each term in it is wrong and to “cure” it we must add counter-terms of the same (wrong) structure. The fundamental constants are just fitting (bare) parameters. And their values change from order to order. It gives the same renormalized solutions but it is the only consolation. The logic and the mathematics are not satisfactory. No wonder all that does not help out in the non renormalizable theories, for example, in quantum gravity.

Many years ago (in 1980-ies) I encountered divergent perturbative corrections in some simple problem. The perturbation “potential” V(z) was proportional to the Dirac delta-function squared. The matrix elements were clearly divergent. (It is similar to the QED divergences seen in the coordinate representation (distribution products).) I could have made a “regularization” and perform the renormalization procedure but fortunately I found the physical reason of appearing such a singular potential and I managed to reformulate the whole problem in terms of a better initial approximation. The perturbation series became finite from the very beginning, as it should be. I could compare my perturbative results with the exact solution since such was the problem. I published my results in academic journals of the former USSR.

Since then I have been sure that the problems in QED are of the same nature – due to too bad initial approximation. Finally I found a better approximation and reformulated the theory in the way that no divergences appear. I answered all questions that puzzle people in QED. In order to explain the mathematical and physical reasons of appearing the perturbative corrections to the fundamental and known constants, I reduced the problem to its essential part and uploaded it to arxiv. There you can find the answer why the renormalizations “work”. I hope I made the story sufficiently transparent so that any person knowing the classical mechanics and quantum mechanics can easily find the answers. You will see why corrections to the fundamental constants appear, when they are legitimate and when they are not, and how to reformulate the theory of interacting things in a correct way.

Briefly, in QED the distribution product in my approach is not singular because of natural electron form-factor: the electron charge is not point-like any more in my initial approximation but smeared quantum mechanically due to coupling to the quantized electromagnetic field. The electron and the quantized EMF form together an “electronium” whose mathematical structure is similar to the atomic one. We all know that the negative and positive (!) charges in an atom are not point-like but smeared. I called the corresponding compound system in QED an “electronium”. In other words, I managed to take the coupling exactly in the zeroth order (initial) approximation of the field equation solutions. I think it is the most physical approach to describing interacting fields. The theory now reminds a potential interaction of compound systems ("fermioniums") in the usual quantum mechanics.

For those who do not get bored with reading simple articles, I provide the references:
“Reformulation instead of Renormalizations”, arxiv:0811.4416
and “Atom as a “Dressed” Nucleus”, arxiv:0806.2635 by Vladimir Kalitvianski. The latter was published in the Central European Journal of Physics, V. 7, N. 1., pp. 1-11, (2009) (Russian publications are not available in the West).

You will not regret reading them. I am very interested in discussing my ideas and solutions.
I am also interested in getting a position in academia to be able to finalize my research.

I thank you all for your participation in this thread.

Cordially,

Bob.

 

[Bob_for_short]

Dear Strangerep,

Thank you for your letter. I received it as an e-mail. Surprisingly it is not visible in the forum!

Thank you for your suggestions on my writing style. It is not perfect, I admit, and it is not my purpose to offend the people. My style is explained partially by the lack of “good education” and partially by the lack of feedback from scientists who could make suggestions at the early stage of writing. Any suggestions are welcome anytime.

Concerning the “classical” rather than “quantum” consideration, I made it on purpose. Historically first I worked out that in QM and then reduced it to a simple CM problem to make the problem the most explicit. If you take patience to finish reading my papers despite my bad stile, you will find it justified, I hope. CM approach is the simplest, comprehensible, and unambiguous. It reveals the problem of correction appearing in the most transparent way. Transition to QM is then elementary.

My “dressing” is somewhat different from the approaches known from the literature. I hope you will catch the difference quickly. My dressing is an ansatz (insight) based on the atomic analogy. In 1985 I found that any atom has a positive charge form-factor and inelastic scattering channels under scattering at large angles, and I quickly realized that it was the physical and technical solution of QED conceptual problems. I published it first as a preprint of Sukhumi Institute of Physics and Technology, then as an article in the Ukrainian Physical Journal, and recently as an article in the CEJP (2009). You will see that the problem of infrared and ultraviolet divergences is eliminated at one stroke by using a better, more correct physically and closer to the exact solutions mathematically initial approximation for the solutions.

Please, do not take my bad style too personally. Just try to extract the useful information.

Sincerely yours,

Bob.

 

[strangerep]

Dear Strangerep,
Thank you for your letter. I received it as an e-mail. Surprisingly it is not visible in the forum!

I changed my mind and decided that some of my suggestions were not appropriate
in a public forum, so I deleted my posting. But you're obviously subscribed to this thread
so you got it immediately by email before I deleted it.

Concerning the “classical” rather than “quantum” consideration, I made it on purpose. Historically first I worked out that in QM and then reduced it to a simple CM problem to make the problem the most explicit. If you take patience to finish reading my papers despite my bad stile, you will find it justified, I hope. CM approach is the simplest, comprehensible, and unambiguous. It reveals the problem of correction appearing in the most transparent way. Transition to QM is then elementary.

My “dressing” is somewhat different from the approaches known from the literature.
I hope you will catch the difference quickly.

Hmmm. Your paper arxiv:0811.4416 is 26 pages, most of which is classical non-relativistic
motivation for your "trial relativistic Hamiltonian" for "novel QED" in your eq(60) on p25.
Your Hamiltonian contains a free part in terms of what you call "electronium" or "fermionium",
expressed in terms of center-of-inertia (CI) momentum P, together with an interaction
potential expressed in terms of CI coordinates R and modified relative coordinates
r which involve both R and the electric field E.

In other approaches, sometimes called "unitary dressing transformation", one tries to
deduce a better Hamiltonian in terms of composite fields (so-called "dressed particles",
although I don't like that term), using certain criteria about the desired new Hamiltonian
to guide one's choice of transformation. (References: M.I.Shirokov in arxiv:nucl-th/0102037,
E.V.Stefanovich in arxiv:hep-th/0503076, and other references therein.)

These other approaches differ from yours in that they start from standard relativistic
QED/QFT, and then deduce perturbatively what the "correct" modified Hamiltonian should
be (but only up to low order). In contrast, you use classical non-relativistic arguments to
motivate a new QED Hamiltonian eq(60) as an ansatz. Apparently, you claim this is the
exactly correct QED Hamiltonian? I.e., not merely a low order approximation?
If so, such a claim is not substantiated in your paper. You only say that "no infrared
divergences arise" and "[...]No ultraviolet divergences arise either [...]", but without
giving details. To substantiate such a large claim requires much more extensive treatment,
quantizing the theory of which eq(60) is a starting point, showing that it is
relativistically acceptable in every way, deriving scattering formulae for well-known
cases, including the anomalous magnetic moment to high order, all done in a rigorous way
(not merely qualitivatively). If such treatment and calculations already exist, they should be
translated into English and put on the arxiv.

You might also consider discussing your ideas on the Independent Research forum,
https://www.physicsforums.com/forumdisplay.php?f=146
which is where some of this discussion belongs, in my humble opinion.
(If you decide to do this, be sure to read the guidelines at the top of that forum,
explaining the necessary format of the postings there.)

 

[neo86]

I think you somewhat misunderstood why the need for renormalization occurs, there's no physical reality behind this rather strange looking process (but the effects you get via renormalization group, e.g. running couplings are of course real), the basic problems lies in the fact that in the theory there are things which are mathematically not rigorously defined and this is what haunts one in QFT at the point where renormalization comes in. To be exact: The Propagators are distributions not functions (this is obvious because they are Green's functions for some differential operator). Now when you start to include quantum effects, by going beyond tree level in perturbation theory (i.e. start calculating loop diagrams) you start to multiply propagators, but since these are distributions you aren't allowed to multiply them and because of this you start to encounter all these weird infinities. Renormalization is just a way to take care of this problem, but if you would have been more careful with the math in the first place you wouldn't have encoutered these infinities at all. The problem with working mathematically rigorously is, that this is really involved, goes quite deep into advanced math and it's much more complicated to get the same results as you get in the "sloppy" way.
The mathematical careful way goes under the name of "causal perturbation theory".
So basically renormalization is just a way to make sense out of the multiplication of objects you aren't supposed to multiply.

Another thing:

Lorentz invariance, gauge invariance, classical limits, CPT, spin-statistics, propagator poles at physical parameters,

A few of these you don't actually put into your QFT but they follow out of QFT and very basic assumptions: namely CPT and spin-statistics.
Spin statistics follows essentially out of looking at the homotopy groups of the 4-d Lorentz group and some quite involved technicalities.
CPT follows out of the fact that your theory should be lorentz invariant,causal and local and the fact that the Hamiltonian of the theory should be bounded from below. (you need this to get a vacuum anyways)Neo

 

[Bob_for_short]

I think you somewhat misunderstood why the need for renormalization occurs, there's no physical reality behind this rather strange looking process (but the effects you get via renormalization group, e.g. running couplings are of course real), the basic problems lies in the fact that in the theory there are things which are mathematically not rigorously defined and this is what haunts one in QFT at the point where renormalization comes in. To be exact: The Propagators are distributions not functions (this is obvious because they are Green's functions for some differential operator). Now when you start to include quantum effects, by going beyond tree level in perturbation theory (i.e. start calculating loop diagrams) you start to multiply propagators, but since these are distributions you aren't allowed to multiply them and because of this you start to encounter all these weird infinities. Renormalization is just a way to take care of this problem, but if you would have been more careful with the math in the first place you wouldn't have encoutered these infinities at all. The problem with working mathematically rigorously is, that this is really involved, goes quite deep into advanced math and it's much more complicated to get the same results as you get in the "sloppy" way.
The mathematical careful way goes under the name of "causal perturbation theory".
So basically renormalization is just a way to make sense out of the multiplication of objects you aren't supposed to multiply.

Dear Neo86,

Thank you for your kind explanation, really. For your information, I know this explanation since my studentship (1975-1981). I had a book of N.N. Bogolubov and Shirkov with this approach including R-operation.

As I explained in one of my previous postings, I encountered very similar problem in the early 80-ies and I first though to "determine" the value of a singular integral from the experimental data. Very soon I realized to what extent this idea was naive and wrong. Fortunately I managed to reformulate the problem and to get rid of the singular integrand in a natural way. And I know everything about Green's functions, I am Green's function master.

Concerning distribution products, in my approach there is a natural regularization factor that makes the perturbative corrections vanish or small, depending on case.

If renormalizations were just a way to make sense out of the multiplication of objects "you aren't supposed to multiply" (?), then this would work without fail in the non-renormalizable theories, like quantum gravity. Renormalizations (=correction discarding) are not legitimate. It may work and it may not work, depending on chance. Please find an explanation in my paper.

Another thing:
A few of these you don't actually put into your QFT but they follow out of QFT and very basic assumptions: namely CPT and spin-statistics.
Spin statistics follows essentially out of looking at the homotopy groups of the 4-d Lorentz group and some quite involved technicalities.
CPT follows out of the fact that your theory should be lorentz invariant,causal and local and the fact that the Hamiltonian of the theory should be bounded from below. (you need this to get a vacuum anyways)

Nice! My approach includes all that but does not use so singular interaction Hamiltonian. That is why no regularization and renormalizations are necessary. Everything makes sense in my theory, in particular, phenomenological (or fundamental) masses and charges.

Regards,

Bob.

 

[Bob_for_short]

I changed my mind and decided that some of my suggestions were not appropriate in a public forum, so I deleted my posting. But you're obviously subscribed to this thread so you got it immediately by email before I deleted it.

Dear Strangerep,

Thank you for your second letter and for being a real gentleman.

Your paper arxiv:0811.4416 is 26 pages, most of which is classical non-relativistic
motivation for your "trial relativistic Hamiltonian" for "novel QED" in your eq(60) on p25.
Your Hamiltonian contains a free part in terms of what you call "electronium" or "fermionium",
expressed in terms of center-of-inertia (CI) momentum P, together with an interaction
potential expressed in terms of CI coordinates R and modified relative coordinates
r which involve both R and the electric field E.

In other approaches, sometimes called "unitary dressing transformation", one tries to
deduce a better Hamiltonian in terms of composite fields (so-called "dressed particles",
although I don't like that term), using certain criteria about the desired new Hamiltonian
to guide one's choice of transformation. (References: M.I.Shirokov in arxiv:nucl-th/0102037,
E.V.Stefanovich in arxiv:hep-th/0503076, and other references therein.)

These other approaches differ from yours in that they start from standard relativistic
QED/QFT, and then deduce perturbatively what the "correct" modified Hamiltonian should
be (but only up to low order). In contrast, you use classical non-relativistic arguments to
motivate a new QED Hamiltonian eq(60) as an ansatz. Apparently, you claim this is the
exactly correct QED Hamiltonian? I.e., not merely a low order approximation?
If so, such a claim is not substantiated in your paper. You only say that "no infrared
divergences arise" and "[...]No ultraviolet divergences arise either [...]", but without
giving details. To substantiate such a large claim requires much more extensive treatment,
quantizing the theory of which eq(60) is a starting point, showing that it is
relativistically acceptable in every way, deriving scattering formulae for well-known
cases, including the anomalous magnetic moment to high order, all done in a rigorous way
(not merely qualitivatively). If such treatment and calculations already exist, they should be
translated into English and put on the arxiv.

You might also consider discussing your ideas on the Independent Research forum,
https://www.physicsforums.com/forumdisplay.php?f=146
which is where some of this discussion belongs, in my humble opinion.
(If you decide to do this, be sure to read the guidelines at the top of that forum,
explaining the necessary format of the postings there.)

Yes, you are right, and I already submitted a thread there. No reply so far. They take time to make a decision.

I agree with you that in order to substantiate my claims it is necessary to carry out a series of specific calculations in a rigorous way.

Below I would like to answer your questions.

The other "dressed" particle approaches deal indeed with the standard QED/QFT Hamiltonians (I am aware of them, believe me) and that is their main drawback. They are bound to perform the renormalizations in one or another way since it is intrinsic feature of the theory. We cannot get rid of infinities just by the variable change. We need another Hamiltonian and better initial approximations from the very beginning. That is why the other “dressed” particle approaches stall. (The electron in my “electronium” is really dressed.)

As R. Feynman said, we are still in the equation guessing stage. In "Reformulation instead of Renormalizations" I showed (on a simple example) that the correct Lagrangian (or Hamiltonian) cannot be derived from or reduced to the "standard" (self-interacting) Lagrangian because here we deal with a "welded" systems (fermioniums in gauge QFT). We have to build (or guess) a Novel QED Hamiltonian from some physical and mathematical judgments rather than just to quantize the wrong Classical Electrodynamics. In my work I have been inspired by the atomic example as an appropriate physical picture (see “Atom as a “Dressed” Nucleus” attached to this posting). The paper "Reformulation instead of Renormalizations" was written later - to reveal the true cause of appearing corrections to the fundamental constants. I repeat: the latter was reduced to the classical mechanical problem on purpose to make things explicit. (Talking about distributions or Green’s functions is not an explanation. It is just a tautology: we obtain divergent values because the integrals diverge.)

I think my ansatz is sufficiently grounded: it is reasonable from physical and mathematical points of view. I presented the idea and its implementation together with some non-relativistic estimations.

First of all, I do not have the infrared problem. I am bound to work with inelastic processes and with the inclusive picture because otherwise I obtain zero (no elastic scattering). It corresponds to the experimental situation. Even from good sense, agree, it is better to take the highly probable processes into account rather than to neglect them and have a divergence problem later on.

The electron (and positron) charge is smeared quantum mechanically, just like the positive charge in an atom (described with a form-factor). This gives a natural regularization factor and makes the calculations sensible mathematically. In free fermionium scattering there is no contribution to the fundamental constants. In bound state problems there are some small corrections that usually considered as correction to the energy levels.

The Lamb shift and the anomalous magnetic moment estimations give very reasonable non-relativistic results, first obtained by T. Welton. I only outline them and refer to his paper to make my presentation shorter. In particular, the anomalous magnetic moment estimation gives correct value but the opposite sign because in the non-relativistic approach one neglects the contribution of the “negative frequency” solution components. (It is know/shown that they give similar contribution.)

No divergences appear on this way. All that is already a good sign even in the non-relativistic approach.

There should not be problems in the relativistic calculations. Relativistic covariance of QED in the Coulomb gauge (or better, in terms of the gauge invariant Dirac variables) has been proven. Briefly, many useful things are just taken from the usual QED (Fock spaces, Hilbert space, etc.). The main difference is in better initial approximations and in a better (potential) interaction Hamiltonian. No getting out of the Hilbert space happens due to a too singular (self-) interaction Hamiltonian (shown by P. Dirac to happen in the standard QED).

I invite you to read both of my papers because it is essential to have in mind the flawless atomic example as a model.

With best regards,

Bob.

 

[strangerep]

I invite you to read both of my papers because it is essential to have in mind
a flawless atomic example as a model.
Bob.

Your paper "Atom_CEJP.pdf" doesn't really tell me much more. You still
reach a similar place near the end where you postulate a "relativistic"
H_QED in eq(23), followed by a brief paragraph of insufficiently-justified
statements.

But I have a question: in your paper arxiv:0811.4416, you write down a
"non-relativistic QED" Hamiltonian in eq(54) where the V interaction
term involves a sum over electric field modes E_{k,\lambda}
up to a "k_max = m_e c/hbar". So apparently, you're imposing a
cutoff. But in similar expressions later, including your "trial
relativistic Hamiltonian of the Novel QED" eq(60), and the following
paragraph, you don't explicitly state the upper limit of the sum over k.
Is the upper limit infinity, or are you still imposing a cutoff like k_max?

 

[zetafunction]

my question is , is there a link between the renormalization and the zeta regularization of infinities ? [tex] 1+2^r +3^r +...... = \zeta (-r) [/tex]

 

[Bob_for_short]

Your paper "Atom_CEJP.pdf" doesn't really tell me much more. You still reach a similar place near the end where you postulate a "relativistic" H_QED in eq(23), followed by a brief paragraph of insufficiently-justified statements.

I do not postulate but propose or advance my point of view. It means flexibility, if you like.

The purpose of this article is twofold: to present a good (working without fail) model of taking into account exactly the "vacuum field fluctuations" and then to propose a Novel QED Hamiltonian basing on the atomic example insight. I really hope that the "atomic" part of the article is not skipped by reader but studied with a pencil.

My ("insufficiently justified") statements are correct, this is the main point. I had no place to dive in details in the frame of one article. Yet I explained in words why it is so (form-factor influence). To demonstrate this, the non-relativistic cross section calculation is sufficiently detailed in it and its physics is quite eloquent. The relativistic calculation gives similar cross section properties. I agree that what is evident to me, may not be so evident to a fresh reader with a different physical picture in mind.

But I have a question: in your paper arxiv:0811.4416, you write down a "non-relativistic QED" Hamiltonian in eq(54) where the V interaction term involves a sum over electric field modes E_{k,\lambda} up to a "k_max = m_e c/hbar". So apparently, you're imposing a cutoff. But in similar expressions later, including your "trial relativistic Hamiltonian of the Novel QED" eq(60), and the following paragraph, you don't explicitly state the upper limit of the sum over k. Is the upper limit infinity, or are you still imposing a cutoff like k_max?

In the exact relativistic Hamiltonian there is no cutoff. All photon frequencies contribute. The trick is that in the exact relativistic approach there also contributions from "negative frequency" solution components that "cancel" (modify) essentially the high frequency oscillator contributions. This is the exact theory result. In the non-relativistic case this property can be preserved and reasonably reduced to a sum over finite range of photon momenta or frequency.

I have not published the concrete relativistic calculation for many reasons.

Thank you for your discussions, I really appreciate them. I am interested in further discussions. Please, feel free to clarify any subtleties, any motivations, etc.

With best regards,

Bob.

 

[strangerep]

I have not published the concrete relativistic calculation for many reasons.

Thank you for your discussions, I really appreciate them.
I am interested in further discussions.

The details of the full relativistic case are what interests me most, so
probably I won't have much more to say until after you publish that.

 

[Bob_for_short]

The details of the full relativistic case are what interests me most, so probably I won't have much more to say until after you publish that.

Well, thank you anyway. I am glad that I have not heard any conceptual objections from participants.



Bob.

 

[Bob_for_short]

The details of the full relativistic case are what interests me most, so probably I won't have much more to say until after you publish that.

The Novel QED relativistic calculations differ from the standard QED calculations: the interaction Hamiltonian describes a four-fermion potential scattering of compound fermioniums, so the calculation technique is different with naturally taking into account multi-particle electronium structure from the very beginning. It means, for example, inclusive calculations: summation over final states as well as averaging over the initial states since one cannot prepare an electronium in its ground state (in the scattering problems). In bound state description there is averaging over oscillator influence, etc. This is a separate subject and it deserves a separate introduction and treatment.

Regards,

Bob.

 

[Bob_for_short]

For copying-and-pasting:

π²³ ∞ 0° ~ µ ρ σ ∑ Ω √ ∫ ≤ ≥ ± ∃ … ⋅ θ φ ψ ω Ω α β γ δ ∂ ∆ ∇ ε λ Λ Γ ô