2

I saw in the book of Mark Hamilton "Mathematical of gauge theory" in the pag. 126 that the killing form to lie algebra $\mathfrak{u}(2)$ is

$$B(X, Y) = 4 Tr(XY) - 2 Tr(X)Tr(Y).$$ I want prove this for this case (not the general case $\mathfrak{gl}(2, \mathbb{C})$). I am trying take the basis $(T^0, T^j) := (i Id, i\sigma_j)$, where $\sigma_j$ are Pauli's matrices, then, by one hand, a matrix $A \in \mathfrak{u}(2)$ can be write as

$$A = \frac{Tr(A)}{2i} T^0 + a_j T^j.$$ So,

$$ad(A) = \sum_{j=1}^{3} a_j ad(T^j),$$ because $ad(T^0)=0$. But, when I use the definition

$$B(A,B) := Tr(ad(A)ad(B)) = \sum_{j=1}^{3} \sum_{l=1}^{3} a_j b_l Tr(ad(T^j)ad(T^l)),$$ I don't get progress, and seem to me that the $Tr(A)$ won't appear because the $T^0$ componente don't appear. Can anyone help me?

Appreciate.

glS
  • 6,818
  • 1
    You don't want to use the general case of $\mathfrak{gl}(2)$? Anyway, here is the proof (or this post). – Dietrich Burde Jun 04 '21 at 14:57
  • the formula you give is for $\mathfrak{gl}(2,\mathbb R)$, not $\mathfrak{u}(2)$. See e.g. https://en.wikipedia.org/wiki/Killing_form – glS Jun 08 '21 at 13:05
  • @DietrichBurde i don't understand the connection between the formula for $\mathfrak{gl}(\mathbb{R})$ and $\mathfrak{u}(n)$.. which one has the formula in the question? i believe if an algebra a is a subalgebra of b, a inherits b's killing form, is that correct? but is this the case here? we have different fields, no? thanks a lot for any clarification. cheers from Italy – l4teLearner May 23 '22 at 22:38

1 Answers1

2

Denote with $\{t_1,t_2,t_3\}$ a basis for $\mathfrak{su}(2)$ such that $[t_i,t_j]=\epsilon_{ijk}t_k$ (implicit sums over repeated indices will be assumed whenever suitable; I'll also not distinguish between lower and upper indices). Note that, in terms of the Pauli matrices, these are $t_i=-\frac{i}{2}\sigma_i$.

  1. (First method) The Killing form is defined as $B(X,Y)={\rm Tr}({\rm ad}(X)\circ{\rm ad}(Y))$, for $X,Y\in\mathfrak{su}(2)$ with decompositions $X=X_i t_i,Y=Y_i t_i$.

    We have $$({\rm ad}(t_i)\circ{\rm ad}(t_j))(t_k) = [ t_i, [t_j, t_k]] = \sum_{\ell m}\epsilon_{jk\ell} \epsilon_{i\ell m} t_m = \sum_{\ell m}\epsilon_{jk\ell} \epsilon_{m i\ell} t_m. $$ We can now use the identity $\epsilon_{jk\ell} \epsilon_{m i\ell} = \delta_{jm}\delta_{jk} - \delta_{ji}\delta_{km}$, to obtain $$({\rm ad}(t_i)\circ{\rm ad}(t_j))(t_k) = \delta_{ki} t_j - \delta_{ji} t_k. $$ For the trace, you need to take the $t_k$ term in this expression, and then sum over $k$: $${\rm Tr}({\rm ad}(t_i)\circ{\rm ad}(t_j)) = \sum_k (\delta_{ki}\delta_{jk} - \delta_{ji}) = -2 \delta_{ij},$$ and thus $$B(X,Y)\equiv {\rm Tr}({\rm ad}(X)\circ{\rm ad}(Y)) = \sum_{ij} X_i Y_j {\rm Tr}({\rm ad}(t_i)\circ{\rm ad}(t_j)) = -2\sum_{i} X_i Y_i. $$ If you represent $X,Y$ as matrices in the standard way, then $${\rm Tr}(XY) = \sum_{ij} X_i Y_j \left(-\frac{1}{4}{\rm Tr}(\sigma_i \sigma_j)\right) = -\frac{1}{2}\sum_{i} X_i Y_i. $$ We conclude that $$B(X,Y) = 4{\rm Tr}(XY).$$

  2. (Another method) We can also leverage the general expression of the Killing form in terms of the structure coefficients. If $[t_i,t_j]=\sum_k {c_{ij}}^k t_k$, then $B(t_i,t_j)=\sum_{mn} {c_{im}}^n {c_{jn}}^m$. In the case of ${c_{ij}}^k=\epsilon_{ijk}$, as is the case for $\mathfrak{su}(2)$, we get $$B(t_i,t_j) = \sum_{mn} \epsilon_{imn}\epsilon_{jnm} = - \sum_{mn} \epsilon_{imn}\epsilon_{jmn} = -2\delta_{ij},$$ and the rest follows as before.

A more general discussion about deriving the Killing forms for the classical Lie algebra is found in How to derive the general formula for the Killing form for classical Lie algebras?.

glS
  • 6,818
  • 1
    hi, is this the formula for $\mathfrak{u}$ (as for the question) or $\mathfrak{su}$? also, what is the connection between the Killing formula for $\mathfrak{u}$, $\mathfrak{gl}(n,\mathbb{R})$ and $\mathfrak{gl}(n,\mathbb{C})$? thanks a lot, I am looking at this and other related question and I am a bit confused. I have also left a comment in the question. – l4teLearner May 24 '22 at 05:50