Skip to main content
just pointed out that a,b are diagonal
Source Link

Building on the paper Idel and Wolf - Sinkhorn normal form for unitary matrices Colin McQuillan suggested, it is easy to see that every $\operatorname{SU}(N)$ matrix $m$ can be decomposed as

$$ m = a \,u\, b $$ with $a,b \in \operatorname{SU}(N)$ and diagonal, and
$$ u = F \,\left(\frac{1}{\mathrm{det}(V)} \oplus V\right) F^\dagger \in \operatorname{SU}(N) $$

where $F_{jk} = \frac{1}{\sqrt{N}}e^{\frac{2\pi i}{N}jk}$ is the Fourier transform unitary matrix and $V\in \operatorname{U}(N-1)$.

Notice that the number of parameters of this factorization matches the dimension of $\operatorname{SU}(N)$.

Building on the paper Idel and Wolf - Sinkhorn normal form for unitary matrices Colin McQuillan suggested, it is easy to see that every $\operatorname{SU}(N)$ matrix $m$ can be decomposed as

$$ m = a \,u\, b $$ with $a,b \in \operatorname{SU}(N)$ and
$$ u = F \,\left(\frac{1}{\mathrm{det}(V)} \oplus V\right) F^\dagger \in \operatorname{SU}(N) $$

where $F_{jk} = \frac{1}{\sqrt{N}}e^{\frac{2\pi i}{N}jk}$ is the Fourier transform unitary matrix and $V\in \operatorname{U}(N-1)$.

Notice that the number of parameters of this factorization matches the dimension of $\operatorname{SU}(N)$.

Building on the paper Idel and Wolf - Sinkhorn normal form for unitary matrices Colin McQuillan suggested, it is easy to see that every $\operatorname{SU}(N)$ matrix $m$ can be decomposed as

$$ m = a \,u\, b $$ with $a,b \in \operatorname{SU}(N)$ and diagonal, and
$$ u = F \,\left(\frac{1}{\mathrm{det}(V)} \oplus V\right) F^\dagger \in \operatorname{SU}(N) $$

where $F_{jk} = \frac{1}{\sqrt{N}}e^{\frac{2\pi i}{N}jk}$ is the Fourier transform unitary matrix and $V\in \operatorname{U}(N-1)$.

Notice that the number of parameters of this factorization matches the dimension of $\operatorname{SU}(N)$.

Link to suggestion, and name of paper
Source Link
LSpice
  • 12.9k
  • 4
  • 45
  • 69

Building on the paper Idel and Wolf - Sinkhorn normal form for unitary matrices Colin McQuillan suggestedsuggested, it is easy to see that every $\mathrm{SU}(N)$$\operatorname{SU}(N)$ matrix $m$ can be decomposed as

$$ m = a \,u\, b $$ with $a,b \in \mathrm{SU}(N)$$a,b \in \operatorname{SU}(N)$ and
$$ u = F \,\left(\frac{1}{\mathrm{det}(V)} \oplus V\right) F^\dagger \in \mathrm{SU}(N) $$$$ u = F \,\left(\frac{1}{\mathrm{det}(V)} \oplus V\right) F^\dagger \in \operatorname{SU}(N) $$

where $F_{jk} = \frac{1}{\sqrt{N}}e^{\frac{2\pi i}{N}jk}$ is the Fourier transform unitary matrix and $V\in \mathrm{U}(N-1)$$V\in \operatorname{U}(N-1)$.

Notice that the number of parameters of this factorization matches the dimension of $\mathrm{SU}(N)$ $\operatorname{SU}(N)$.

Building on the paper Colin McQuillan suggested, it is easy to see that every $\mathrm{SU}(N)$ matrix $m$ can be decomposed as

$$ m = a \,u\, b $$ with $a,b \in \mathrm{SU}(N)$ and
$$ u = F \,\left(\frac{1}{\mathrm{det}(V)} \oplus V\right) F^\dagger \in \mathrm{SU}(N) $$

where $F_{jk} = \frac{1}{\sqrt{N}}e^{\frac{2\pi i}{N}jk}$ is the Fourier transform unitary matrix and $V\in \mathrm{U}(N-1)$.

Notice that the number of parameters of this factorization matches the dimension of $\mathrm{SU}(N)$ .

Building on the paper Idel and Wolf - Sinkhorn normal form for unitary matrices Colin McQuillan suggested, it is easy to see that every $\operatorname{SU}(N)$ matrix $m$ can be decomposed as

$$ m = a \,u\, b $$ with $a,b \in \operatorname{SU}(N)$ and
$$ u = F \,\left(\frac{1}{\mathrm{det}(V)} \oplus V\right) F^\dagger \in \operatorname{SU}(N) $$

where $F_{jk} = \frac{1}{\sqrt{N}}e^{\frac{2\pi i}{N}jk}$ is the Fourier transform unitary matrix and $V\in \operatorname{U}(N-1)$.

Notice that the number of parameters of this factorization matches the dimension of $\operatorname{SU}(N)$.

Source Link

Building on the paper Colin McQuillan suggested, it is easy to see that every $\mathrm{SU}(N)$ matrix $m$ can be decomposed as

$$ m = a \,u\, b $$ with $a,b \in \mathrm{SU}(N)$ and
$$ u = F \,\left(\frac{1}{\mathrm{det}(V)} \oplus V\right) F^\dagger \in \mathrm{SU}(N) $$

where $F_{jk} = \frac{1}{\sqrt{N}}e^{\frac{2\pi i}{N}jk}$ is the Fourier transform unitary matrix and $V\in \mathrm{U}(N-1)$.

Notice that the number of parameters of this factorization matches the dimension of $\mathrm{SU}(N)$ .