https://www.kitp.ucsb.edu/statement-by-the-yangmills-working-group
The proof is claimed to be "rigorous", which is strange, because in quantum mechanics nothing seems to be rigorous.
What is the momentum p of the energy E in the vacuum? Any value given to the momentum p introduces preferred coordinates and breaks Galilei covariance
The proof is based on a lattice model in which we assume that "zero-point fluctuations" of the fields are allowed if the wavelength of the fluctuation is larger than the lattice spacing.
There is a lot of energy E in zero-point fluctuations. In some continuum models, the energy density is infinite. The momentum p of that energy is zero relative to the lattice. But this breaks Galilean covariance: all inertial frames should be equivalent in vacuum.
How to restore Galilean covariance? "Renormalize" the energy density of the zero-point fluctuations to zero by subtracting some constant C from the density. If the energy density is zero, then we have no need to assign any velocity, or momentum, to that density.
We declare the lowest energy of the zero-point fluctuations as the "vacuum". An example from the real world: let us have a crystal. Even at zero kelvin, there is oscillation of atoms and groups of atoms in the crystal. They are the zero-point fluctuations in this system. We declare the state at 0 K as the "vacuum" in that system and define its energy density to be zero.
Let us then reduce the lattice spacing toward zero and take the limit. We keep the energy density of zero-point fluctuations renormalized to zero. This is what the OpenAI proof presumably does.
However, there is a problem in this. If we have interactions between fields, then matter can interact with zero-point fluctuations in the vacuum. Those interactions can extract some energy from the zero-point fluctuations, like they do if the Casimir effect is explained through zero-point fluctuations.
The problem: what is the momentum p of the energy E which we extract from the vacuum? We may assume that the momentum is zero relative to the lattice. But then we break Galilean invariance because the frame of the lattice becomes preferred.
Did OpenAI prove that extracting energy from the vacuum is impossible? We will check that.
If energy can be extracted from the vacuum then we also break energy conditions used in general relativity. Those conditions require that energy density cannot be negative.
Any "interaction" with the vacuum breaks Galilean covariance?
What does an "interaction" mean?
Suppose that a material system of particles can extract energy from the vacuum, and the amount of that energy depends on the configuration of the particles. This is clearly an "interaction".
Or could we define it in this way: the system of particles emits a negative energy "field", depending on the configuration of particles?
However, a negative energy "field" is not something which lattice quantum field theory has.
Rudolf Haag's theorem
Rudolf Haag's theorem claims that if there is vacuum polarization, then the standard framework of quantum field theory encounters a serious problem. The problem does not appear in non-interacting (trivial) quantum field theories.
We have to check if Haag's theorem is equivalent to our own complaint about broken Galilei covariance.
On the Internet some people claim that lattice quantum field theory somehow circumvents Haag's theorem. However, the limit of OpenAI's lattices no longer is a lattice – it is a continuum quantum field theory.
Lattice gauge theory can restore Galilean covariance in the continuum limit? The Casimir effect
metal plates
| |
| |
| |
reduced
vacuum
energy
Suppose that we have static metal plates which extract energy from vacuum fluctuations through the Casimir effect when we slowly let the plates approach each other. The energy E extracted from the vacuum has a zero momentum p = 0 in the laboratory frame.
Let us switch to a frame which moves at a constant velocity v relative to the laboratory. In that frame it might be the that the plates extract vacuum energy which moves at the same speed v? Then Galilean covariance would hold.
The interaction of matter fields with the vacuum is very problematic
Even if we renormalize the vacuum energy density to zero, vacuum fluctuations still interact with matter fields. The interaction necessarily changes individual vacuum fluctuation modes. Otherwise, there would not be an interaction. How does the OpenAI model handle these enormously complex, and probably infinite, interactions? The interactions transfer an infinite stream of information to the vacuum!
The model looks hopeless: we cannot "renormalize" away infinitely complex interactions.
In our blog we have repeatedly remarked that assuming that the vacuum is not empty leads to insurmountable problems in a model.
In the case of the Casimir effect, the energy pulling the plates together must come from electromagnetic vacuum fluctuations. Moving two electrons in such a vacuum would involved a very complex process of interaction with vacuum fluctuations. How does the OpenAI model prove that we can ignore those interactions?
*** WORK IN PROGRESS ***