Site Map Polytopes Dynkin Diagrams Vertex Figures, etc. Incidence Matrices Index

General Unitary Matrices

Matrices with complex entries admit 2 fundamental operations,

The result of the composition of both these transformations is called the hermitian matrix. Matrices are defined to be unitary, iff their inverses are given by their hermitians. - Sure any purely real valued unitary matrix is therefore an orthogonal matrix. The set of unitary n x n-matrices is denoted by U(n).


Unitary 2 x 2-matrices

The general member of U(2) is easily derived. It is given by

e
(
 a  b 
-b* a*
)

where further |e| = 1 and |a|2 + |b|2 = 1.

Proof

Let A, B, C, D be non-negative reals (absolut values) and α, β, γ, δ be phases (arguments).

(
1  0
0  1
)
=
(
Ae  Be
Ce  De
)
(
Ae-iα  Ce-iγ
Be-iβ  De-iδ
)
=
(
A2 + B2         ACei(α-γ) + BDei(β-δ)
ACe-i(α-γ) + BDe-i(β-δ)        C2 + D2
)

Case A=0: From entry (1,1) it thus follows B=1. From (1,2) or (2,1) it then follows D=0. From (2,2) it then follows C=1. For simplicity assume further α = δ = 0 as well. Therefore:

(
Ae  Be
Ce  De
)
= ei(β+γ+π)/2
(
 0     ei(β-γ-π)/2
-e-i(β-γ-π)/2    0
)

Case B=0: From (1,1) it thus follows A=1. From (1,2) or (2,1) it then follows C=0. From (2,2) it then follows D=1. For simplicity assume further β = γ = 0 as well. Therefore:

(
Ae  Be
Ce  De
)
= ei(α+δ)/2
(
ei(α-δ)/2     0
0    e-i(α-δ)/2
)

Cases C=0 or D=0 are identical to the formers.

Else: We thus can assume hence forward that A, B, C, D are strictely positive.
From (1,2) and (2,1) we get -AC/BD = ei(-α+β+γ-δ) = -1.
From (1,1) and (2,2) we can assume that A=cos(φ), B=sin(φ), C=sin(ψ), D=cos(ψ) with some 0<φ<π/2 and 0<ψ<π/2. Together this gives -tan(ψ)/tan(φ)=-1, that is φ=ψ (in this range). So in fact we have A=D, B=C.
Further we have α-β=γ-δ+π. Therefore:

(
Ae  Be
Ce  De
)
= ei(α+δ)/2
(
Aei(α-δ)/2     Bei(-α+2β-δ)/2
Bei(-α+-δ)/2    Ae-i(α-δ)/2
)
= ei(α+δ)/2
(
 Aei(α-δ)/2      Bei(-α+2β-δ)/2
-Be-i(-α+-δ)/2    Ae-i(α-δ)/2
)


Low-order unitary 2 x 2-matrices

Matrices u which follow the equation un = 1 (for the first time) are said to be of order n.

The case of order 1 is trivial, u clearly is the identity matrix 1.

Order 2 unitary 2 x 2-matrices u have 2 general possibilities, eiter u = -1 or

±
(
v   c 
c* -v
)

where further v is real, subject to v2 ≤ 1, and c is subject to |c|2 = 1 - v2.

Proof
(
1  0
0  1
)
= e
(
 a  b 
-b* a*
)
e
(
 a  b 
-b* a*
)
= e2
(
 a2-bb*   (a+a*)b 
-(a+a*)b*  a*2-bb*
)

From entry (1,2) or (2,1) we have either b=0 or a+a*=0.

Case b=0: Then |a|=1 and further from (1,1) and (2,2) e2a2=e2a*2=1. Thence 0=e2(a2-a*2)=e2(a+a*)(a-a*). Cause |e|=|a|=1, hence either a is purely imaginary or purely real, i.e. a=ik for some integral k.

Subcase k even (a real): Here we have e2=1 or e=±1. But equal signs of a and e give order 1 only. Thus:

(
-1  0
 0 -1
)

Subcase k odd (a imaginary): Here we have e2=-1 or e=±i. Extracting the further factor i in front of the matrix we get:

±
(
1  0
0 -1
)

Case a+a*=0 (a imaginary): Here we can use the fact that uu=1=uu*, because we both assume order 2 and unitarity. Thus we have especially u=u*, or from entry (1,1) of u itself: ea=e*a*=e*(-a) (the latter identity by assumption of pure imaginarity of a). Therefore either a=0 or e=±i

Subcase a=0: Then |b|=1. u=u* restricted to (1,2) further gives eb=-e*b and thus leads to e=±i in this case as well. So we are left with:

±
(
0   (ib)
(ib)*  0
)

Else: Neither a nor b is zero, but a=iv for some real number v and e=±i.

e
(
 a  b 
-b* a*
)
= ±i
(
 iv   b 
-b* -iv
)
= ±
(
-v   ib 
(ib)* v
)

Order 3 (ζ3 = exp(2πi/3)): unitary 2 x 2-matrices u of this order again have 2 general possibilities, either

(
ζ3m  0
0   ζ3n
)

but not both m ≡ 0 and n ≡ 0 (modulo 3) – in which case u would be the identity matrix, which clearly is of order 1 only –, or

ζ3k
(
-1/2 + iv   b 
-b*        -1/2 - iv
)

where further v is real, subject to v2 ≤ 3/4, and b only is subject to bb* = 1 - aa* = 3/4 - v2.



© 2004-2024
top of page