[guided]The purpose of this step is to show that clutching maps are not extra structure: they are exactly what remains after the bundle has been trivialized on the two hemispheres.
Let
\begin{align*}
\pi: E \to S^n
\end{align*}
be a rank $k$ vector bundle. The restrictions to the two closed hemispheres are trivial for a specific reason: $D_+^n$ and $D_-^n$ are compact Hausdorff, hence paracompact, and each is contractible because it is homeomorphic to the closed disk. The triviality theorem for vector bundles over contractible paracompact bases therefore applies to $E|_{D_+^n}$ and $E|_{D_-^n}$. Choose explicit trivializations $\Phi_+: E|_{D_+^n} \to D_+^n \times F$ and $\Phi_-: E|_{D_-^n} \to D_-^n \times F$. For $x\in D_+^n$, write $\Phi_+(x):E_x\to F$ for the fibre coordinate map determined by $\Phi_+$; for $x\in D_-^n$, write $\Phi_-(x):E_x\to F$ for the fibre coordinate map determined by $\Phi_-$.
For a point $x \in S^{n-1}$, the fibre $E_x$ is seen in two coordinate systems: one coming from $\Phi_+(x)$ and one coming from $\Phi_-(x)$. Passing from the $+$ coordinates to the $-$ coordinates gives the linear isomorphism
\begin{align*}
g_{\Phi_+,\Phi_-}(x) := \Phi_-(x)\circ \Phi_+(x)^{-1}: F \to F.
\end{align*}
The map $g_{\Phi_+,\Phi_-}: S^{n-1} \to G$ is continuous because it is the transition function between the two vector bundle charts. More explicitly, after writing each fibre map in the fixed [vector space](/page/Vector%20Space) $F$, composition and inversion are continuous operations in the topological group $G$, so $x \mapsto \Phi_-(x)\circ \Phi_+(x)^{-1}$ is continuous.
Now compare $E$ with the glued bundle $E_{g_{\Phi_+,\Phi_-}}$. On $D_+^n$ use $\Phi_+$, and on $D_-^n$ use $\Phi_-$. If $x \in S^{n-1}$ and a vector has $+$-coordinate $v \in F$, then its $-$-coordinate is $g_{\Phi_+,\Phi_-}(x)v$. This is exactly the relation
\begin{align*}
(x,v)_+ \sim (x,g_{\Phi_+,\Phi_-}(x)v)_-.
\end{align*}
Define a map from the disjoint union $(D_+^n\times F)\sqcup(D_-^n\times F)$ to $E$ by sending $(x,v)_+$ to $\Phi_+^{-1}(x,v)$ and sending $(x,w)_-$ to $\Phi_-^{-1}(x,w)$.
If $(x,v)_+ \sim (x,g_{\Phi_+,\Phi_-}(x)v)_-$, then the definition of $g_{\Phi_+,\Phi_-}$ gives
\begin{align*}
g_{\Phi_+,\Phi_-}(x)v=\Phi_-(x)\Phi_+(x)^{-1}v.
\end{align*}
Hence
\begin{align*}
\Phi_-^{-1}(x,g_{\Phi_+,\Phi_-}(x)v)=\Phi_+^{-1}(x,v).
\end{align*}
Therefore the map is constant on equivalence classes and descends to a well-defined bundle map $E_{g_{\Phi_+,\Phi_-}}\to E$. It is fibrewise linear and bijective because, over each point of $S^n$, either one product chart applies or the two representatives are identified by the displayed relation. Its inverse is obtained by applying $\Phi_+$ or $\Phi_-$ according to the hemisphere representative, so the descended map is a vector bundle isomorphism
\begin{align*}
E_{g_{\Phi_+,\Phi_-}} \cong E.
\end{align*}[/guided]