- #1
Assaf
- 2
- 0
"Handedness" of Mobius transformations
Hello everyone,
I've been trying to derive the SU(2) (right-handed) rotation matrix by using a projection of the sphere and Mobius transformations, but I'm having some issues which I was hoping someone here could help me out with. My apologies for the long question, but I couldn't find any way to make it shorter. :)
What I'm trying to re-derive is the SU(2) 2x2 complex rotation matrix appearing in Sakurai's book "Modern Quantum Mechanics", on page 166 of the revised (red) edition. I unfortunately can't seem to get the Latex to work on my end, but the same matrix also appears, say, in these lecture notes (p. 10):
http://www.physics.rutgers.edu/grad/502/Lectures_Final/Lec04_SU2.pdf"
I've started by looking at the stereographic projection of points on the sphere onto the complex plane: the projection P' of P=(Px,Py,Pz) (P is on the sphere) is the intersection of the line NP with the xy-plane, where N is the sphere's north pole (N=(0,0,1)). This leads to the following coordinates for P':
P' = (Px/(1-Pz), Py/(1-Pz), 0)
Since I want to identify the xy-plane with the complex plane, I set the correspondence
P' <--> z = Px+i*Py = Pxy/(1-Pz)
where Pxy = Px+i*Py.
So far so good. Here comes the puzzling bit. To deduce the form of the 2x2 rotation matrix about z, I start by applying a z-rotation to a vector in xyz. Such a rotation is represented by the "regular" real 3x3 rotation matrices, as found here:
http://en.wikipedia.org/wiki/Rotation_matrix
(To the best of my knowledge my choice of handedness is consistent: the rotation matrices shown here, both real and complex, are right-handed. Please correct me if I'm wrong.)
So, under a z-rotation, the vector P=(Px,Py,Pz) transforms into
P ---> Prot = (Pxrot, Pyrot, Pzrot) = (Px*cos(phi) - Py*sin(phi), Px*sin(phi) + Py*cos(phi), Pz)
The corresponding complex number, obtained by stereographic projection, transforms as:
z = (Px + iPy)/(1-Pz) --> (Pxrot+i*Pyrot)/(1-Pzrot) = e^{i*phi}*z = exp(i*phi/2)*z / (exp(-i*phi/2))
So, this is a Mobius transformation of the form M(z) = (a*z+b)/(c*z+d) with ab-cd=1, a=exp(i*phi/2), d = exp(-i*phi/2), b=c=0. The corresponding matrix is then:
exp(i*phi/2) 0
0 exp(-i*phi/2)
But this is the exact opposite (in terms of handedness) of the matrices appearing in Sakurai
i.e., that's a left-handed rotation. Can the two views be brought into agreement? And if not, why?
Hello everyone,
I've been trying to derive the SU(2) (right-handed) rotation matrix by using a projection of the sphere and Mobius transformations, but I'm having some issues which I was hoping someone here could help me out with. My apologies for the long question, but I couldn't find any way to make it shorter. :)
What I'm trying to re-derive is the SU(2) 2x2 complex rotation matrix appearing in Sakurai's book "Modern Quantum Mechanics", on page 166 of the revised (red) edition. I unfortunately can't seem to get the Latex to work on my end, but the same matrix also appears, say, in these lecture notes (p. 10):
http://www.physics.rutgers.edu/grad/502/Lectures_Final/Lec04_SU2.pdf"
I've started by looking at the stereographic projection of points on the sphere onto the complex plane: the projection P' of P=(Px,Py,Pz) (P is on the sphere) is the intersection of the line NP with the xy-plane, where N is the sphere's north pole (N=(0,0,1)). This leads to the following coordinates for P':
P' = (Px/(1-Pz), Py/(1-Pz), 0)
Since I want to identify the xy-plane with the complex plane, I set the correspondence
P' <--> z = Px+i*Py = Pxy/(1-Pz)
where Pxy = Px+i*Py.
So far so good. Here comes the puzzling bit. To deduce the form of the 2x2 rotation matrix about z, I start by applying a z-rotation to a vector in xyz. Such a rotation is represented by the "regular" real 3x3 rotation matrices, as found here:
http://en.wikipedia.org/wiki/Rotation_matrix
(To the best of my knowledge my choice of handedness is consistent: the rotation matrices shown here, both real and complex, are right-handed. Please correct me if I'm wrong.)
So, under a z-rotation, the vector P=(Px,Py,Pz) transforms into
P ---> Prot = (Pxrot, Pyrot, Pzrot) = (Px*cos(phi) - Py*sin(phi), Px*sin(phi) + Py*cos(phi), Pz)
The corresponding complex number, obtained by stereographic projection, transforms as:
z = (Px + iPy)/(1-Pz) --> (Pxrot+i*Pyrot)/(1-Pzrot) = e^{i*phi}*z = exp(i*phi/2)*z / (exp(-i*phi/2))
So, this is a Mobius transformation of the form M(z) = (a*z+b)/(c*z+d) with ab-cd=1, a=exp(i*phi/2), d = exp(-i*phi/2), b=c=0. The corresponding matrix is then:
exp(i*phi/2) 0
0 exp(-i*phi/2)
But this is the exact opposite (in terms of handedness) of the matrices appearing in Sakurai
i.e., that's a left-handed rotation. Can the two views be brought into agreement? And if not, why?
Last edited by a moderator: