I am reading Prof. Eduardo Fradkin's QFT lecture note (the link to the lecture note is gone). In his note, he considered the Wilson loop operator
ˆWΓ(x,x)=ˆP(expig∮Γ(x,x)dzμAμ(z))
where Γ(x,x) denotes a loop with both its initial and final points at x, g is the coupling constant, and Aμ is the non-Abelian gauge field. We want to obtain the definition of the field tensor Fμν by expanding WΓ(x,x) to the leading order when we have infinitesimal Γ(x,x). This means the length of the loop Γ(x,x) is infinitesimal and also the oriented surface whose boundary is Γ(x,x) has its minimal area infinitesimal. We then proceed as follows
ˆWΓ(x,x)≈ˆP(1+ig∮Γ(x,x)dzμAμ(z)+12[ig∮Γ(x,x)dzμAμ(z)]2+⋯).
The first two terms are easy to do via Stoke's theorem:
ˆP(1)=1ˆP(ig∮Γ(x,x)dzμAμ(z))=ig∮Γ(x,x)dzμAμ(z)=ig2∫Σdxμ∧dxν(∂μAν−∂νAμ).
My question is that he then directly wrote down
ˆP(12[∮Γ(x,x)dzμAμ(z)]2)=12∫Σdxμ∧dxν(−[Aμ,Aν]).
**I am a bit confused about how can we turn the path-ordered product of two line integrals into something like a surface integral.** I currently do not have an idea about where to start to prove this identity. But it seems to be crucial if we want to prove that, for infinitesimal loop Γ(x,x), we have
ˆWΓ(x,x)≈1+ig2∫Σdxμ∧dxνFμν
where
Fμν=∂μAν−∂νAμ−ig[Aμ,Aν].
Are there any suggestions or good reference that carry out this computation
ˆP(12[∮Γ(x,x)dzμAμ(z)]2)=12∫Σdxμ∧dxν(−[Aμ,Aν])
in details? Thanks!