I'm writing the answer to my question but I'm not completely sure though
Intro
What do we want to show ?
$$U(n) \simeq \frac{SU(n) \times U(1)}{\mathbb{Z}_{n}}$$
Groups definition
- $U(n)$ = the group of $n\times n$ unitary matrices $\Rightarrow$ $U \in U(n): UU^{\dagger} = U^{\dagger}U = I \Rightarrow \mid det (U) \mid ^{2} = 1$
- $U(1) =$ the group of $1\times 1$ unitary matrices $\Rightarrow U(1) = \lbrace e^{i\varphi} \mid \varphi \in \left[ 0, 2\pi \right] \rbrace$
- $SU(n) =$ the group of $n\times n$ unitary matrices with determinant 1
- $\mathbb{Z}_{n} =$ the cyclic group of n integers modulo n $\Rightarrow$ $\mathbb{Z}_{n} =$ the set $\lbrace 0,1,2,...,n-1 \rbrace$ with the operation of addition modulo n.
\end{itemize}
Isomorphism theorem
We use the first isomorphism theorem:
Let G and H be groups and let $f: G \longrightarrow G'$ be a group homomorphism. Then:
$$Im f \simeq \frac{G}{Ker f} $$
What do we have to do ?
We have to do the following things:
- Find a map $f: SU(n) \times U(1) \rightarrow U(n)$ and show that it's a homomorphism
- Show that $Ker f = \mathbb{Z}_{n}$
- Show that $Im f = U(n)$ which is equivalent to show that f is surjective
We find and homomorphism between $SU(n) \times U(1)$ and $U(n)$
\begin{eqnarray*}
f: && SU(n) \times U(1) \rightarrow U(n) \\
&& (S,e^{i\varphi}) \mapsto Se^{i\varphi}
\end{eqnarray*}
$f$ is a homomorphism if :
\begin{eqnarray*}
f((S_{1},e^{i\varphi _{1}})(S_{2},e^{i\varphi _{2}} )) = f(S_{1},e^{i\varphi _{1}})f(S_{2},e^{i\varphi _{2}} )&& \forall \varphi _{1}, \varphi _{2} \in U(1) \\
&& \forall S_{1},S_{2} \in SU(n)
\end{eqnarray*}
It's easy to show:
\begin{eqnarray*}
f((S_{1},e^{i\varphi _{1}})(S_{2},e^{i\varphi _{2}})) &=& f(S_{1}S_{2},e^{i\varphi _{1}}e^{i\varphi _{2}} ) \\
&=& S_{1}S_{2} e^{i\varphi _{1}} e^{i\varphi _{2}} \\
&=& e^{i\varphi _{1}} S_{1} e^{i\varphi _{2}} S_{2} \\
&=& f(S_{1},e^{i\varphi _{1}})f(S_{2},e^{i\varphi _{2}})
\end{eqnarray*}
We show that $Ker f = \mathbb{Z}_{n}$
$$Ker f = \lbrace (S,e^{i\varphi}) \in (SU(n) \times U(1)) \: \vert \: f(S,e^{i\varphi}) = Se^{i\varphi} = e_{U(n)} = I \rbrace$$
Let's find S
\begin{eqnarray*}
e^{i\varphi}S&=& I \\
S &=& e^{-i\varphi} I\\
det(S) &=& det(e^{-i\varphi} I) \\
1 &=& e^{-in \varphi} \\
1 &=& \cos (n\varphi ) - i \sin (n \varphi)
\end{eqnarray*}
$$
\left\{
\begin{array}{r c l}
\cos (n \varphi ) &=& 1\\
\sin (n \varphi ) &=& 0\\
\end{array}
\right.
\Rightarrow n \varphi = 2k \pi \Rightarrow \varphi = \frac{2k \pi}{n} \quad \text{with} \quad k \in \mathbb{Z}$$
So the matrices S are:
$$S = e^{-i \frac{2 k \pi}{n}} I \quad \text{with} \quad k \in \lbrace 0,1,...,n-1 \rbrace$$
Let's find $e^{i\varphi}$
\begin{eqnarray*}
e^{i\varphi}S&=& I \\
e^{i\varphi}e^{-i \frac{2 k \pi}{n}} I&=& I \\
e^{i\varphi}&=& e^{i \frac{2 k \pi}{n}} \\
\end{eqnarray*}
Ker f
$$Ker f = (e^{-i \frac{2 k \pi}{n}} I,e^{i \frac{2 k \pi}{n}}) \quad \text{with} \quad k \in \lbrace 0,1,...,n-1 \rbrace $$
We show that Ker $f$ is isomorph to $\mathbb{Z}_{n}$
We denote by $\mathbb{Z}_{n}$ the cyclic group of integers modulo n.
Let's not forget that:
The group law $\cdot$ of Ker f is given by:
$(S_{1},e^{i\varphi _{1}})\cdot(S_{2},e^{i\varphi _{2}}) = (S_{1}S_{2},e^{i(\varphi _{1} + \varphi _{2})})$
The group law $\cdot$ of $\mathbb{Z}_{n}$ is given by:
$k_{1} \cdot k_{2} = k_{1} + k_{2}$
\end{itemize}
Let's show that Ker f is isomorph to $\mathbb{Z}_{n}$
$$\phi: Ker f = (S,e^{i\varphi}) \rightarrow \mathbb{Z}_{n} : (e^{-i \frac{2 k \pi}{n}} I,e^{i \frac{2 k \pi}{n}}) \mapsto k$$
$\phi$ is an homomorphism
$$\phi((S_{1},e^{i\varphi _{1}})(S_{2},e^{i\varphi _{2}})) = \phi(S_{1},e^{i\varphi _{1}}) + \phi(S_{2},e^{i\varphi _{2}})$$
\begin{eqnarray*}
\phi((S_{1},e^{i\varphi _{1}})(S_{2},e^{i\varphi _{2}}))&=& \phi(S_{1}S_{2},e^{i(\varphi _{1} + \varphi _{2})}) \\
&=& \phi (e^{-i \frac{2 (k_{1} + k_{2}) \pi}{n}} I,e^{i \frac{2 (k_{1} + k_{2}) \pi}{n}}) \\
&=& k_{1} + k_{2} \\
\phi(S_{1},e^{i\varphi _{1}}) + \phi(S_{2},e^{i\varphi _{2}})&=& (e^{-i \frac{2 k_{1} \pi}{n}} I,e^{i \frac{2 k_{1} \pi}{n}}) + (e^{-i \frac{2 k_{2} \pi}{n}} I,e^{i \frac{2 k_{2} \pi}{n}}) \\
&=& k_{1} + k_{2}
\end{eqnarray*}
$\phi$ is injective
- $\phi$ is injective $\Leftrightarrow \forall \varphi_{1},\varphi_{2} \in \left[ 0, 2\pi \right] \: \text{and} \: \forall S_{1},S_{2} \in SU(n): \phi (S_{1},e^{i\varphi _{1}}) = \phi (S_{2},e^{i\varphi _{2}}) \Rightarrow \varphi_{1} = \varphi_{2} \: \text{and} \: S_{1} = S_{2}$
\begin{eqnarray*}
\phi (S_{1},e^{i\varphi _{1}}) & = & \phi (S_{2},e^{i\varphi _{2}})\\
S_{1}e^{i\varphi _{1}} & = & S_{2}e^{i\varphi _{2}} \\
e^{i(\varphi _{1} - \varphi _{2})} S_{1} & = & S_{2} \\
det( e^{i(\varphi _{1} - \varphi _{2})} S_{1}) & = & det( S_{2} ) \\
e^{in(\varphi _{1} - \varphi _{2})} & = & 1
\end{eqnarray*}
$$\Rightarrow \varphi _{1} = \varphi _{2} \Rightarrow S_{1} = S_{2}$$
- We can also use the fact that $\phi$ is injective if and only if $Ker \phi = \lbrace e_{Ker f} \rbrace = \lbrace (I,1)\rbrace$
$$Ker \phi = \lbrace (e^{-i \frac{2 k \pi}{n}} I,e^{i \frac{2 k \pi}{n}}) \in Ker f \: \vert \: \phi(e^{-i \frac{2 k \pi}{n}} I,e^{i \frac{2 k \pi}{n}}) = e_{\mathbb{Z}_{n}} = 0 \rbrace$$
$$\phi(e^{-i \frac{2 k \pi}{n}} I,e^{i \frac{2 k \pi}{n}}) = k = 0 \Rightarrow Ker \phi = (e^{0} I,e^{0}) = (I,1)$$
$\phi$ is surjective
$\phi$ is injective $\Leftrightarrow \forall k \in \mathbb{Z}_{n}, \exists (e^{i \frac{2 k \pi}{n}} I,e^{-i \frac{2 k \pi}{n}}) \in Ker f : \phi(e^{i \frac{2 k \pi}{n}} I,e^{-i \frac{2 k \pi}{n}}) = k$
Another way to show $\phi$ is surjective is by noting that fact that $\phi$ is an injective map between 2 finites sets with the same number of elements so $\phi$ is surjective
We show that f is surjective
How do we show that $Im f = U(n)$ ?
We want to show that $Im f = U(n)$ where: $ Im f = \lbrace Se^{i\varphi} \vert S \in SU(n), \varphi \in \left[ 0, 2\pi \right] \rbrace$
To do that, we can show that:
$$Im f \subseteq U(n) \quad \text{and} \quad U(n) \subseteq Im f$$
$Im f \subseteq U(n)$
$$Im f \subseteq U(n) \Rightarrow \mid det (Se^{i\varphi}) \mid ^{2} = 1 \quad \forall S \in SU(n), \forall \varphi \in \left[ 0, 2\pi \right] $$
\begin{eqnarray*}
\mid det (Se^{i\varphi}) \mid ^{2} & = & \mid e^{i n \varphi} \mid ^{2} \\
& = & 1 ^{2} \\
& = & 1
\end{eqnarray*}
$U(n) \subseteq Im f$
It's equivalent to show that: $\forall X \in U(n), \exists (Y,z) \in SU(n) \times U(1) \quad \text{such as} \quad f(Y,z) = Yz = X$
\begin{eqnarray*}
X \in U(n)&\Rightarrow& \vert det(X) \vert ^{2} = 1 \Rightarrow det(X) = e^{i\theta} \equiv m \\
&\Rightarrow& X^{\dagger}X = XX^{\dagger} = I \\
\end{eqnarray*}
Let's write X as: $X = zY = m^{1/n} Y$ where $Y = m^{-1/n}X$ and let's show that $Y \in SU(n)$ and $m^{1/n} \in U(1)$:
$Y \in SU(n) \Leftrightarrow YY^{\dagger} = Y^{\dagger} Y = I$ and $det(Y) = 1$
$$YY^{\dagger} = (m^{-1/n} X)(m^{-1/n} X)^{\dagger} = (e^{\frac{i\theta}{n}} X)(e^{\frac{i\theta}{n}} X)^{\dagger} = e^{\frac{i\theta}{n}} e^{\frac{-i\theta}{n}} XX^{\dagger} = I $$
$$det(Y) = det(m^{-1/n} X) = m^{-1} det(X) = m^{-1} m = 1 \Rightarrow Y \in SU(n)$$
- $m^{1/n} = e^{\frac{i\theta}{n}} \in U(1) $
The homomorphism is indeed a rather trivial one.
Consider the following map $\psi : S_n \to \mathbb P_{n \times n}$, where the latter is the set of permutation matrices of order $n$. This map is defined as follows : given $\phi \in S_n$, the $i$th column of $\psi(\phi)$ is the column vector with a $1$ in the $\phi(i)$th position, and $0$ elsewhere.
It is easy to see that $\psi(\phi)$ is indeed a permutation matrix, since a $1$ occurs in any position if and only if that position is described by $(\phi(i),i)$, for any $1 \leq i \leq n$.
This map is clearly multiplicative. The proof is awkward to write, but easy to understand. We must prove that $\psi(\alpha \circ \beta)=\psi(\alpha)\psi(\beta)$ for any permutations $\alpha,\beta$.
Note that since these are matrices, it's enough to show that each corresponding entry is equal. So let us take the entry $(i,j)$ of each matrix.
Then : $\psi(\alpha \circ \beta)_{ij} = 1$ if and only if $i = \alpha \circ \beta(j)$, as I said earlier. Otherwise, it is zero.
Note that by ordinary matrix multiplication, $(\psi(\alpha)\psi(\beta))_{ij} = \sum_{k=1}^n \psi(\alpha)_{ik}\psi(\beta)_{kj}$.
Now, we know that $\psi(\alpha)_{ik} = 1$ only when $i = \alpha(k)$. Similarly, $\psi(\beta)_{kj} = 1$ only when $k = \beta(j)$. Hence, their product is one precisely when both of these happen : $i = \alpha(k),k = \beta(j)$. If both these don't happen simultaneously, then whenever one of $\psi(\alpha)_{ik},\psi(\beta)_{kj}$ is one the other will be zero, so the whole sum will be zero. However, this is the same as saying that the sum is one exactly when $i = \alpha \circ\beta(j)$. This description matches with the decription given for $\psi(\alpha \circ \beta)_{ij}$ given earlier.
Hence, entry by entry these matrices are the same. Therefore the matrices are the same, and hence $\psi$ is a homomorphism between the two spaces, an isomorphism as it has trivial kernel and the sets are of the same cardinality.
Best Answer
Just use the same homomorphism, and then double the angle. Dividing by negatives means that we identify numbers on opposite sides of the origin. Double the angle, and that's not a problem anymore. After all, we can map the circle 2-1 to the circle.
(Or, as Tsemo Aristide's answer put it, going from the projective line to the circle is an angle-doubling)