While preparing this note over the past few days, I had extensive conversations with Grok 4. Most of them were helpful. I was just testing. For instance Grok solved Exercise 2 below. Even if it was easy, it required logical thinking, and that Grok does pretty well. On the other hand, Grok sometimes behaves like a hound dog. You ask it to bring you a duck, and it happily brings you a hen. Then, when you are upset, it apologizes: 'Yes, indeed, that is not exactly what you asked for, and I know it.'
This post is a continuation of The three involutions of Cl(2,2). Our aim is to examine the group Spin(2,2) - the double covering group of SO0(2,2). SO0(2,2) is the connected component of the identity of the group O(2,2) - the conformal group of our toy spacetime with only one space dimension. We need to start with some definitions. The conventions may differ from one author to another, so I am choosing just one, suiting my personal taste at the moment, but mostly I am trying to follow the notation used in D.J.H. Garling, “Clifford Algebras: An Introduction”, London Mathematical Society, CUP 2011. A good introduction to Clifford algebra can be found in the lecture notes “Math 210C: Clifford algebras and spin groups” by Brian Conrad (Stanford University, available as a free PDF at http://math.stanford.edu/~conrad/210CPage/handouts/clifford.pdf).
General Definitions:
The Clifford norm Δ(a)
For any Clifford algebra Cl(V,q) , V-vector space, q - quadratic form on V, we define
1) The quadratic norm Δ(a)
Δ(a) = ν(a)a, a∈Cl(V,q) (1)
2) The Clifford group Γ(V.q)
The group Γ(V,q) is the group of invertible elements g∈Cl(V,q) having the property
α(g) v g-1 ∈ V if v ∈ V. (2)
Here α is the main automorphism: α(x) = x for even elements, α(x) = -x for odd elements of the Clifford algebra. In the previous post we have found the explicit form of α for Cl(2,2) represented as Mat(4,R), but that is not important at the moment.
Exercise 1. Carefully verify that Γ(V,q) is a group.
Exercise 2. Using the properties of α show that if g∈Γ(V.q) , then α(g)∈Γ(V.q).
We will be interested in the group Spin+(V,q) defined as:
Spin+(V,q) = {g ∈ Cl(V,q)+: Δ(g) = 1, g ∈ Γ(V.q) }. (3)
Note: In (3) , the condition Δ(g) = 1, 1 should be understood as the Clifford algebra element 1. In a matrix representation it will be the identity matrix, not a number.
Spin+(2,2) Explicitly
With these foundations in place, let's apply them to our case of Cl(2,2).
Let us now restrict ourselves to the case at hand, and find an explicit form of elements of Spin+(2,2). An element g of Cl(2,2) should belong to the even Clifford algebra Cl(2,2). In our matrix representation Cl(2,2) = Mat(4,R), while Cl+(2,2) consists of block diagonal matrices of the form {{A,0},{0,B}}. Then we need the condition Δ(g) = 1. In The three involutions of Cl(2,2) we have found that
ν(x) = UxTU-1, (4)
where
U = -e1e2. (5)
In our representation generators are given by matrices
e1 = {{0, 0, 0, 1},
{0, 0, 1, 0},
{0, 1, 0, 0},
{1, 0, 0, 0}}
e2 = {{0, 0, 1, 0},
{0, 0, 0, -1},
{1, 0, 0, 0},
{0, -1, 0, 0}}
e3 = {{0, 0, -1, 0},
{0, 0, 0, -1},
{1, 0, 0, 0},
{0, 1, 0, 0}}
e4 = {{0, 0, 0, -1},
{0, 0, 1, 0},
{0, -1, 0, 0},
{1, 0, 0, 0}}
Calculating U we get, in the block matrix form
U = {{ε,0},{0,ε}}, (6)
where ε = {0,1},{-1,0}}.
We can now calculate Δ(g) for g = {{A,0},{0,B}}. The result is the block diagonal matrix
Δ(g) = {{det(A)I2,0},{0,det(B)I2}}. (7)
Computational Verification
For similar symbolic calculation we can use free Maxima software. Here is the Maxima code that does our calculation:
e1: matrix([0, 0, 0, 1], [0, 0, 1, 0], [0, 1, 0, 0], [1, 0, 0, 0]);
e2: matrix([0, 0, 1, 0], [0, 0, 0, -1], [1, 0, 0, 0], [0, -1, 0, 0]);
e3: matrix([0, 0, -1, 0], [0, 0, 0, -1], [1, 0, 0, 0], [0, 1, 0, 0]);
e4: matrix([0, 0, 0, -1], [0, 0, 1, 0], [0, -1, 0, 0], [1, 0, 0, 0]);
ω: e1 . e2 . e3 . e4;
/* Define the 2x2 matrices A and B */
A: matrix([a11, a12], [a21, a22]);
B: matrix([b11, b12], [b21, b22]);
/* Define 2x2 zero matrix */
Z: zeromatrix(2, 2);
/* Create the block matrix u = [[A, 0], [0, B]] */
a: mat_unblocker(matrix([A, Z], [Z, B]));
u: -e1 . e2;
u . transpose(a) . invert(u) . a;
From (7) we see that in the matrices g = {{A,0},{0,B}} in Spin+(2,2) we must have det(A)=det(B)=1.
Exercise 3. Write Maxima code that verifies that for g = {{A,0},{0,B}} with det(A) = det(B) = 1, the third condition g ∈ Γ(V.q) is automatically satisfied. (You can also use online AI to write the code for you!).
Hint: Represent the generators e1–e4 as matrices, compute α(g) v g^{-1} for each basis vector v in V, and check if the result remains in V."
So we have a truly beautiful result:
Spin+(2,2) = { g = {{A,0},{0,B}}: det(A)=det(B)=1}.
Thus we see that, as a group,
Spin+(2,2) = SL(2,R)×SL(2,R).
In the next post, we'll explore the 2:1 homomorphism Spin+(2,2) → SO₀(2,2), followed by an examination of minimal left ideals in Cl(2,2) where spinors reside.
P.S. 15-07-25 11:58 Preparing the next post I was conversing with Grok 4. Here is the Report (written in .tex by Grok himself!) on this conversation:



Yesterday I managed to repeat the calculation of Δ(g) for g = {{A,0},{0,B}}, and proceeded to Ex.3, and obtained for g e1 g^{-1} the following matrix:
ReplyDelete{0, 0, N1, N2
0, 0, N3, N4
-N4, N2, 0, 0
N3, -N1, 0, 0}
where
N1 = a_12 b_22 -- a_11 b_21
N2 = a_11 b_11 -- a_12 b_12
N3 = a_22 b_22 -- a_21 b_21
N4 = a_21 b_11 -- a_22 b_12
and I am not sure that this proves that the "result remains in V".
You then need to show that the obtained matrix is a linear combination of e1,e2,e3,e4.
DeleteYes, that's what I thought. It would be fun to find these coefficients explicitly.
DeleteIndeed, matrix g e1 g^{-1} obtained as a result of transforming basis vector e1
Delete{0, 0, N1, N2
0, 0, N3, N4
-N4, N2, 0, 0
N3, -N1, 0, 0}
has coordinates:
((N2+N3)/2, (N1-N4)/2, -(N1+N4)/2, (N2-N3)/2)
in the basis e1, e2, e3, e4,
where N1, N2, N3, N4 are defined above.
If you look at g e1 g^{-1}, you see that you should obtain an expression that is quadratic in N's.
DeleteYou can also use the following trick for 2x2 matrices: If you assume that det(A)=1, then A^{-1} = -s A^T s, where s={{0,1},{-1,0}} . You need to check the "-" sign, I am not sure
Delete"should obtain an expression that is quadratic in N's"
DeleteNo, N's are combinations of products of elements a_ij and b_ij, probably in this sense they can be considered as 'quadratic'. But I checked straightforwardly that
{0, 0, N1, N2
0, 0, N3, N4
-N4, N2, 0, 0
N3, -N1, 0, 0} =
1/2((N2+N3)e1 + (N1-N4)e2 - (N1+N4)e3 + (N2-N3)e4)
with e1, e2, e3, e4 defined as in the main text of the Post.
OK. Mea culpa. I have forgotten that you have defined N's before. . You should get the same as is in the new blog note that I posted a while ago.
DeleteArk, in (1) letter "v" looks like vector in V and not like "nu" defined by (4), which is a bit misleading
ReplyDeleteYes, it looks similar, but with a magnifying glass you can see the difference!
DeleteIn Exercise 1 the word "carefully" is slightly suspicious. Ark, do you mean that one should carefully consider the cases of odd and even elements?
ReplyDeleteFor even elements, a(g) = g, and three group properties are as follows:
(1) group operation: if v is in V and g1 and g2 are group elements, then
(g1 g2) v (g1 g2)^-1 = g1 (g2 v g2^-1) g1^-1 is evidently also in V;
(2) by definition of group, element g is invertible;
(3) the group has unit element - the Clifford algebra element 1.
In case of odd elements, we have to show that -g v g^-1 is in V, but isn't it true that if g is group element then -g is also group element?