Take a
functor B:C→∫F for which
ΠB=idC. Let us decompose the
action of this
functor on
objects into
functions obj and
elt so that
B(c)=(objB(c),eltB(c))
Then since
ΠB=idC we get that
objB(c)=c for any
c∈C and also that the
action of
B on
morphisms is the identity. Hence the
functor B simply acts to pick out for every
c∈C an element
eltB(c)∈F(c) (and the chosen elements must abide by the coherence condition for the
category of elements).
We may now organize the information contained in
B into a
cone as follows,
defined legwise
αB:1⇒F(αB)c(⋆)=eltB(c)
where
⋆∈1. This definition
satisfies the
cone condition because for any
f:c→d in
C we know since the codomain of
B is the
category of elements ∫F that
F(B(f))(eltB(c))=eltB(d); this fact rephrases to
F(f)∘(αB)c=(αB)d.