Types of quantum field thories
The BPTS construction considers – on a 4-dimensional Minkowski spacetime Wick rotated to the Euclidean – gauge field configurations for gauge group the special unitary group that have vanishing field strength outside some finite radius. These are then equivalently configurations on the 4-sphere. The BPTS instanton is the -gauge field whose underlying -principal bundle has second Chern class=instanton number equal to .
The physics literature typically focuses on describing this -bundle in terms of the Cech cocycle which after covering the 4-sphere with two 4-balls (two “hemispheres”) is given by an -vaued transition function on the intersection of these two balls, which has the homotopy type of the 3-spehere. Since also the manifold underlying the special unitary Lie group is diffeomorphic to , this allows to encode the classes of -principal bundles/-instantons on in terms of homotopy classes of maps , and this is what much of the literature focuses on.
The traditional story in the physics textbooks (copied endlessly from one textbook to the next, over generations) of -instantons (BPTS instantons) tends to fail to highlight some key global points, without which the whole construction really collapses. The following text means to explain the correct description (using the mathmatics of Cech cohomology cocycles via clutching construction on one-point compactification of Minkowski spacetime) but presented and phrased such that the folklore physics story becomes fully visible – including its crucial fix.
For any gauge group (a Lie group) then a -Yang-Mills instanton on some 4-dimensional spacetimes (a pseudo-Riemannian manifold) is a -principal bundle on (this is going to be the “instanton sector”), equipped with a -principal connection (this is the actual gauge field) such that this is self-dual, in that its curvature form satisfies , where denotes the Hodge star operator for the given metric (field of gravity).
A standard theorem says that there is precisely one self-dual principal connection on every isomorphism class of -principal bundles. Therefore classifying and counting instantons amounts to classifying and counting -principal bundles.
This is what makes instantons a “topological” structure in the parlance of physics, meaning that they do not depend on Riemannian metric information, after all.
We now take the spacetime to be Minkowski spacetime and the gauge group to be the special unitary group . This is the case of “BPTS-instantons”. Of course other choices are possible and may lead to richer situations, but this simple case is what the physics textbooks tend to focus on (not the least of course because these choices are relevant for phenomenology of the weak nuclear force as seen in accelerator experiments) and it already serves to highlight key points. Since, as we just said, we may disregard all metric properties, this means now that we regard spacetime to be just the Cartesian space .
Or almost. At this point one needs to be careful with the boundary conditions in order to get the topology right.
Vanishing at infinity
The physical energy condition on an instanton is that the field strength vanishes “at infinity”. Mathematically, a continuous function on a locally compact topological space vanishes at infinity precisely if it extends to the “one-point compactification” of – where we literally adjoin the “point at infinity” “” and glue it to by defining a suitable topology on .
It is this passage from to which takes care of the subtleties that often tend to be glossed over.
There is a classifying space for -principal bundles, denoted . This being a classifying space means that for any paracompact topological space, then the isomorphism classes of -principal bundles on are in bijection with homotopy classes of continuous functions from to .
Now for , then
But this counting of instantons works more generally, if we use a suitable counting function. First of all, there is a topological space whose only homotopy group is , and such that this is the group of integers. This is the Eilenberg-MacLane space :
Now there is a continuous function
to classes in degree-4 cohomology
This cohomology class
But in the special case that the base space is 4-dimensional, then only a single one of these classes may be non-trivial, namely the second Chern class . Therefore this class completely characterizes -principal bundles in 4d.
Constructing instantons from gauge transformations
We may construct the bundles that are classified this way explicitly by using Cech cohomology. This says that we get such a bundle by
choosing transition functions on each double overlap of two charts
such that on triple overlaps the cocycle condition holds.
Now comes the major fact which gives makes this general theory look like the structures that appear in the physics books: the clutching construction.
Namely, for general one needs the charts to form a good open cover in order to guarantee that all isomorphism classes of gauge bundles are captured by the construction via transition functions.
But since everything is topological now, it doesn’t matter that these charts are literally hemi-spheres in the metric sense. In order to get the standard picture we instead make maximally large and take it to cover all of except the “north pole” (which is really the “point at infinity”, due to the one-point compactification above), while we take to be a tiny open neighbourhood of that point, sitting there as a tiny ice cap around the north pole. So
Topologically this is homeomorphic to the situation before, and hence just as good.
So now back to the general prescription of building principal bundles via Cech cohomology, we are to choose transition functions on all overlaps of charts. But thanks to the clutching construction, there is now just a single such overlap, namely
Moreover there is no non-trivial triple overlap, hence no cocycle condition that our transition function it to satisfy.
But notice that that the intersection of two hemi-4-spheres that overlap slightly is just a 3-sphere times a slight thickening:
Moreover, the thickening direction here is trivial, and so one finds that instantons are classified by homotopy classes of maps
This is where this “gauge transformation at infinity” in the physics textbooks really comes from. The key point here is that it is indeed gauge transforming, namely the restriction of our bundle to to its restriction to the “neighbourhood of infinity” .
But as topological spaces and so
So we get the same classification from Cech cohomology that we got from classifying space theory, as it must be. But now we know how that second Chern class may be concretely embodied in that “gauge transformation at infinity”.
Gauge fields vanishing at infinity
Now we bring in connections. As discussed before, we may just as well consider any principal connection. In the Cech cohomology picture and still using the clutching construction this now means to choose
-valued 1-forms on
such that on the tiny intersection of the two charts at infinity we have
Because then the above just says that on becomes gauge trivial “at infinity”, by the given gauge transformation (the one whose winding number counts our instantons).
Counting instantons by integrating
So far we have derived the physics picture of an instanton: An -gauge field which becomes gauge trivial “at infinity”, witnessed by a gauge transformation on the “annulus at infinity” whose winding number is the instanton number. But the key point is that we see that the little neighbourhood of infinity is part of the picture, and that is necessary now to understand the Chern 4-form.
Namely to every -valued 1-form we may assign the ordinary (abelian) 4-form
where is the curvature form of .
But beware now this is only true on a single chart. And just because our chart covers “everything except infinity”, we must not forget that there is a second chart , the “neighbourhood of infinity”.
Namely the 4-form that is defined on the whole of the 4-sphere , this 4-form is only locally exact (as every closed 4-form is, by the Poincare lemma). Generally, we define it chart-wise by
This does indeed give a globally defined 4-form, no matter what the local connection forms are, as long as they satisfy the required condition that they are related by a gauge transformation on the overlap . Because the 4-form is gauge invariant.
So the 4-form thus defined
a potential 3-form when restricted to ;
a potential 3-form when restricted to ;
but it does not have a potential 3-form on all of , unless the instanton number vanishes.
Put this way this should be very obvious now. But it is easy to get confused about this due to the sheer convenience of the clutching construction used above: we actually were allowed to choose !
This might make it superficially look like there is only a single local gauge potential around, and that the 4-form is locally exact. But this is not the case: there are two local gauge potentials on two different charts, and just because one of them happens to be equal to zero still does not mean that the extension of the 4-form to all of the 4-sphere has a global potential. It has a potential only after restricting it to , even if this means “only removing the point at infinity”.
With thus understood that is not globally exact, it becomes believable that the integral
is generally non-vanishing, and is in fact yet another incarnation of the same integer that we had before, the instanton number. That this is so is given to us by Chern-Weil theory.
Outlook: Chern-Simons 2-Gerbes
In fact the full story is nicer still. Namely the local Chern-Simons 3-forms together with the gauge transformation at infinity form a Cech cohomology cocycle for a circle 3-bundle with connection (a bundle 2-gerbe). This is the Chern-Simons 2-gerbe of the gauge field. And the fourth incarnation of the instanton number is: the Dixmier-Douady class of this 2-gerbe
The original articles are
A. A. Belavin, A.M. Polyakov, A.S. Schwartz, Yu.S. Tyupkin, Pseudoparticle solutions of the Yang-Mills equations, Phys. Lett. B 59 (1), 85-87 (1975) doi
A. A. Belavin, V.A. Fateev, A.S. Schwarz, Yu.S. Tyupkin, Quantum fluctuations of multi-instanton solutions, Phys. Lett. B 83 (3-4), 317-320 (1979) doi
For surveys and introductions see the references at Yang-Mills instanton.