In electrodynamics, Coulomb gauge is a simple example of a gauge that gives a one-to-one mapping between field strength and potential. This is not the case in the non-Abelian theory, where the Gribov ambiguity shows that there are configurations of field strengths that allow several distinct choices of Coulomb gauge potentials, as well as other configurations that cannot be brought into Coulomb gauge at all [85, 236]. These issues are often swept under the rug when working in perturbation theory, but cannot be ignored when considering potentials of order 1/g, for which the commutator term in the field strength is comparable in size to the derivative terms. As suggested by Eq. (10.8), it is precisely such potentials with which we will be concerned.
page 201 in Classical Solutions in Quantum Field Theory by Erik Weinberg
In our discussion of gauge fixing and the Fadeev-Popov procedure for functional quantization of gauge theories, we have seen that there could be the problem of Gribov ambiguity. This refers to the fact that there could be different field configurations which obey the same gauge-fixing condition, but which are related by a gauge transformation. This means that we are unable to find a “good gauge fixing” where the gauge-fixing condition chooses one and only one representative configuration for all potentials which are gauge equivalent.
page 334 in Quantum Field Theory: A Modern Perspective by V. P. Nair
There is a difficulty which arises in the cornpactified case which is not present in the more general one. This involves gauge fixing and the Gribov ambiguity. ' As yet the gauge has not been fixed and there is the possibility that fixing the gauge could affect the topology of configuration space. The object of gauge fixing is to choose one member from each class of gauge-equivalent configurations so that there will be no double counting of configurations in the path integral. In the fiber-bundle picture of Dowker, ' this corresponds to finding a section in the bundle, reducing each configuration fiber to a particular configuration in the fiber. The difficulty is that Singer' has proven that there are no global gauges when space is compactified to S . That is, there is no global section of the fiber bundle that Dowker is using. This means that the space of gauge-fixed configurations is disconnected. The answer is that the Gribov ambiguity allows for certain zero-action discontinuous evolutions. Because there are no global sections, the gauge fixing fails to always specify a unique configuration from each configuration fiber. When it does fail, there are two (or more) configurations which are related by a gauge transformation. Each of these lies on a different section of the fiber bundle and one's histories may move between the disconnected sections by way of these configurations.
Nontrivial homotopy and tunneling by topological instantons by Arlen Anderson
It was proven by Singer in "Some Remarks on the Gribov Ambiguity", provided that we compactify spacetime to the 4-sphere, that "no continuous choice of exactly one
connection on each orbit can be made. Thus the Gribov ambiguity for the
Coloumb gauge will occur in all other gauges. No gauge fixing is possible. "
This Gribov ambiguity can be given a precise mathematical characterization in the language of fiber bundles.
[…]
We can think of gauge fixing as follows. When we do gauge fixing, we choose a representative potential A (obeying some gauge fixing condition) for each physical configuration. Thus we are specifying the physical configuration C and a gauge transformation g associated to it which takes it into A. We have an assignment of a point on the fiber, namely, $G_C$ , for each point C in $\mathcal{C}$. In other words, gauge fixing is the choice of a section for the bundle A. If we can choose a section globally, then we have the splitting $A = \mathcal{C} × G_\star$ globally. There is no problem with gauge fixing and no Gribov ambiguity. The existence of the Gribov problem is thus equivalent to the statement that the bundle ($G_∗$ , A, \mathcal{C}$) is nontrivial and does not have a global section. This gives a precise characterization of the Gribov problem as a topological property of the bundle of gauge potentials.
page 335 in Quantum Field Theory: A Modern Perspective by V. P. Nair
The BRST method used to deal with the gauge symmetry of perturbative Yang-Mills theory does not appear to generalize to the full non-perturbative theory, for a rather fundamental reason. This was first pointed out by Neuberger back in 1986 (Phys. Lett. B, 183 (1987), p337-40.), who argued that, non-perturbatively, the phenomenon of Gribov copies implies that expectation values of gauge-invariant observables will vanish.