2

If we know how to write the matrix representation of the fundamental representation of SU(N), could we use them to derive the matrix representation of other representations of SU(N)? (adjoint, anti-symmetric, or symmetric, etc.)

For example, for SU(2), we know the tensor product of 2-dimensional fundamental representation of SU(2) gives the 3-dimensional adjoint representation of SU(2): $$ 2 \times 2 = 3_A + 1_S. $$ So if we know the explicit 2-dimensional matrix representation of SU(2) in terms of its three Lie algebra generators, $$\sigma_1 = \begin{pmatrix} 0&1\\ 1&0 \end{pmatrix}, \quad \sigma_2 = \begin{pmatrix} 0& -i \\ i&0 \end{pmatrix}, \quad \sigma_3 = \begin{pmatrix} 1&0\\ 0&-1 \end{pmatrix} , \tag{1} $$ how do we DERIVE the 3-dimensional matrix representation of SU(2) (which we know is the adjoint representation of SU(2) and vector representation of SO(3))? Namely in terms of its three Lie algebra generators, $$ \boldsymbol{L}_x = \begin{bmatrix}0&0&0\\0&0&-1\\0&1&0\end{bmatrix}, \quad \boldsymbol{L}_y = \begin{bmatrix}0&0&1\\0&0&0\\-1&0&0\end{bmatrix}, \quad \boldsymbol{L}_z = \begin{bmatrix}0&-1&0\\1&0&0\\0&0&0\end{bmatrix}. \tag{2} $$ Namely, how do derive eq (2) from the given eq (1) for the Lie algebra matrix representations?

  • The Lie brackets of matrices are given by commutators, i.e., $[A,B]=AB-BA$. This is given. But then the adjoint representation follows directly by definition. For example, $[x,y]=z$, so that the matrix of $L_x$ has first column zero and second column $(0,0,1)$. – Dietrich Burde Aug 01 '21 at 16:44
  • Dear Dietrich Burde - thanks, but I need to have explicit derivations. I know Lie bracket. I also already know the answer of SU(2) case. How about you illuminate the new question on the SU(5) case? https://math.stackexchange.com/q/4214397/141334 – annie marie cœur Aug 01 '21 at 16:47
  • 1
    The adjoint matrices are exactly given by the Lie brackets, so I don't understand your question. $ad(x)(y)=[x,y]$ by definition. In your example here, $L_x=ad(x)$. – Dietrich Burde Aug 01 '21 at 16:48
  • Indeed, so this adjoint rep is not a good example. How about the other question on SU(5)? there I am not asking the adjoint rep but the other reps. – annie marie cœur Aug 01 '21 at 16:49
  • This question has been answered several times on the physics site, since it is a standard drill of the physics curriculum. The most excruciatingly explicit answer, to the point of overkill, is this one; but I could just jot down the condensed version any good physics theory course requires as an exercise, if you were interested. – Cosmas Zachos Aug 04 '21 at 00:35
  • thanks - this is a trial to fix the language in order to get an explicit answer for https://math.stackexchange.com/q/4214397/141334 -> has the set of 24 matrices of su(5) Lie algebra of rank-10 and rank-15 been given somewhere? – annie marie cœur Aug 04 '21 at 00:43
  • No, not to my knowledge for su(5)… those are huge sparse matrices… a project. – Cosmas Zachos Aug 04 '21 at 09:58

1 Answers1

3

Question: "Namely, how do derive eq (2) from the given eq (1) for the Lie algebra matrix representations?"

Answer: Let $k$ be the real numbers and $K$ the complex numbers and let $G:=SU(2)$ be the special unitary group. It follows its complexification $G_K\cong SL(2,K)$ is the special linear group. The irreducible representations of $SL(2,K)$ are classified. Let $V:=K\{e_1,e_2\}$ be the standard representation of $SL(2,K)$, with dual $V^*:=K\{x_1,x_2\}$. There is a canonical map of $SL(2,K)$-modules

$$ \wedge: V\otimes_K V \rightarrow \wedge^2 V \cong T$$

and since $T$ is the trivial $SL(2,K)$-module it follows $V \cong V^*$ is an isomorphism of $SL(2,K)$-modules. The irreducible $SL(2,K)$-modules are the symmetric powers

$$Sym^d(V) \cong Sym^d(V^*)$$

for $d \geq 0$. "Using" the isomorphism $G_K \cong SL(2,K)$ it follows the irreducible $SL(2,K)$-modules $Sym^d(V),Sym^d(V^*)$ are irreducible $G$-modules with dimension given by

$$dim_K(Sym^d(V))=d+1, dim_k(Sym^d(V))=2(d+1).$$

There are precise relations between irreducible $G$-modules and the irreducible $SL(2,K)$-modules (similar for $SU(3)$ and $SL(3,K)$). The irreducible representations of $SL(n,K)$ have been classified using the "Schur-Weyl" construction. Given any "partition" $\lambda$, there is a Schur-Weyl functor $\mathbb{S}_{\lambda}$ and when you apply this functor to the standard representation

$$V(\lambda):=\mathbb{S}_{\lambda}(V)$$

of $SL(n,K)$ you get all finite dimensional irreducible $SL(n,K)$-modules.

You can find some information in

Hall, Brian - Lie Groups, Lie Algebras, and Representations: An Elementary Introduction (Graduate Texts in Mathematics, 222)

and

Fulton/Harris - Representation Theory: A First Course.

Your comment: "For example, for SU(2), we know the tensor product of 2-dimensional fundamental representation of SU(2) gives the 3-dimensional adjoint representation of SU(2):"

Answer: If $dim_k(V)=2$ it follows $dim_k(V^{\otimes_k n}) = 2^n$. You cannot tensor a 2-dimensional representation with itself and get an odd dimensional representation. The representation $V\otimes V$ may decompose into a direct sum $V_3 \oplus V_1$ with $dim_k(V_i)=i$: The tensor product of two irreducible representations is not necessarily irreducible.

Groups such as $SL(n,K)$ and $SU(n)$ are semi simple Lie groups and they have the property that any finite dimensional $k$ or $K$ representation $V$ decompose into a direct sum

$$V \cong \oplus_i V_i$$

where $V_i$ is an irreducible module. The irreducible modules are classified by their "highest weight vector" and "highest weight". Hence if you want to check if two irreducible modules are isomorphic you must calculate their weights and compare. See the above books for more details.

Example: In your case you have given a basis for $\mathfrak{g}:=\mathfrak{su}(2,k):=k\{\sigma_i\}$

and an explicit representation

$$\rho: \mathfrak{g} \rightarrow End_k(k^2)$$

This induce an explicit adjoint representation

$$ad: \mathfrak{g} \rightarrow End_k(\mathfrak{g})$$

defined by $ad(x)(y):=[x,y]$. Since $dim_k(\mathfrak{g})=3$ you will get explicit matrices $ad(\sigma_i)$ for $i=1,2,3$, and you must calculate these matrices in your given basis $B$. If $x:=\sigma_1$ you must calculate the vectors

$$v_i:=[x, \sigma_i]$$

\begin{align*} [v_i]_B:=\begin{pmatrix} v_{1i} \\ v_{2i} \\ v_{3i} \end{pmatrix} \end{align*}

and construct the matrix

\begin{align*} M_1:= \begin{pmatrix} 0 & v_{12} & v_{13} \\ 0 & v_{22} & v_{23} \\ 0 & v_{32} & v_{33} \end{pmatrix} \end{align*}

Do this for $\sigma_2, \sigma_3$, construct $M_2,M_3$ and compare to the matrices you write down above. You must first check that the basis you give above is a basis for $\mathfrak{su}(2,k)$ as $k$-vector space.

MichaelNgelo
  • 553
  • 3
  • 15
hm2020
  • 10,015
  • thanks this is great - I voted +1 - how about the other one https://math.stackexchange.com/q/4214397/141334 – annie marie cœur Aug 01 '21 at 18:01
  • About this "If $dim_k(V)=2$ it follows $dim_k(V^{\otimes_k n}) =2^n$. You cannot tensor a 2-dimensional representation with itself and get an odd dimensional representation." --> but you can write $2×2=3_+1_$, and hopefully get the $3_$ out of $2×2$ – annie marie cœur Aug 01 '21 at 18:04
  • @anniemariecœur - the same argument applies: You should study the relation between $SU(n)$-modules and $SL(n,K)$-modules. The irreducible $SL(n,K)$-modules are explicitly constructed in Fulton/Harris using $\mathbb{S}_{\lambda}$. – hm2020 Aug 01 '21 at 18:04
  • Thanks - but do you know that really can lead to explicit matrix representations of the dimension I asked? or do you know any refs with the matrix given explicitly? – annie marie cœur Aug 01 '21 at 18:05
  • "There are precise relations between irreducible -modules and the irreducible (2,)-modules" what exactly do you define modules here? – annie marie cœur Aug 01 '21 at 22:59
  • I will be 100% happy if there is way to give a precise set of 24 rank-10 and rank-15 matrix in this math.stackexchange.com/q/4214397/141334 question - so to prove that we do know what we are doing... thanks! – annie marie cœur Aug 01 '21 at 22:59
  • @anniemariecœur - it seems your question is a "linear algebra" question and that you want to express the operators $ad(\sigma_i)$ in a basis for $End(\mathfrak{su}(2,k))$. You must first check that $\sigma_i$ is a basis for $\mathfrak{su}(2,k)$ and then calculate the Lie products. – hm2020 Aug 02 '21 at 12:49
  • @anniemariecœur - it seems $[\sigma_1,\sigma_2]=2i\sigma_3$ - is the above given "basis" a basis for $\mathfrak{su}(2,k)$? – hm2020 Aug 03 '21 at 08:30
  • thanks so much for the update! – annie marie cœur Aug 03 '21 at 15:33
  • "The irreducible modules are classified by their "highest weight vector" and "highest weight". Hence if you want to check if two irreducible modules are isomorphic you must calculate their weights and compare." ---> are the "highest weight vector" and "highest weight vector" necessary and sufficient conditions to distinguish the irreducible modules?

    – annie marie cœur Aug 03 '21 at 15:36
  • Could you also post an answer here math.stackexchange.com/q/4214397/141334 since no one so far shows the explicit rank-10 and rank-15 matrix Lie algebra generators for the su(5), with 24 matrices. – annie marie cœur Aug 03 '21 at 15:40
  • @anniemariecœur - the calculation is similar - you calculate the Lie products of the matrices. I do not think this site is suitable for such lengthy calculatins with matrices. – hm2020 Aug 04 '21 at 08:36
  • @anniemariecœur - "are the "highest weight vector" and "highest weight vector" necessary and sufficient conditions to distinguish the irreducible modules?" You find this topic explained in the above books. The highest weight of an irreducible module classifies the module. Two irreducible modules have the same highest weight iff they are isomorphic. – hm2020 Aug 04 '21 at 09:17
  • could you explain is this your notation "(2,)"? – annie marie cœur Aug 09 '21 at 13:47
  • is that (2,) as (2) acts on -dimensional vector space? thanks! – annie marie cœur Aug 09 '21 at 13:48
  • Is it possible to get an explicit expression for $1,2,3$ in $[]$? Is that just $\pm 1$ or 0 for these coefficients? – annie marie cœur Aug 09 '21 at 13:50
  • Can you also write $M_2:= \begin{pmatrix} v_{11} & 0 & v_{13} \ v_{21} & 0 & v_{23} \ v_{31} & 0 & v_{33} \end{pmatrix}$, $3:= \begin{pmatrix} v{11} & v_{12} & 0 \ v_{21} & v_{22} & 0 \ v_{31} & v_{32} &0 \end{pmatrix}$? But if the basis vectors are the same, I do not see the matrices of yours match mine. – annie marie cœur Aug 09 '21 at 13:54
  • @anniemariecœur - if $SU(n)$ is the special unitary group, we use $\mathfrak{su}(n)$ to denote its Lie algebra. There is a canonical representation $ad:\mathfrak{su}(n) \rightarrow End_k(\mathfrak{su}(n))$ and you must calculate this representation. This calculation involves constructing a basis for $\mathfrak{su}(n)$ and calculating Lie products of matrices. – hm2020 Aug 09 '21 at 17:21
  • can you recommend a section of a book, or a ref PDF to read about this construction? – annie marie cœur Aug 09 '21 at 17:59
  • @anniemariecœur - I give two references above. The books contain many elementary examples. – hm2020 Aug 10 '21 at 08:45
  • dear all, maybe you know this https://math.stackexchange.com/q/4220583/955245 – Марина Marina S Aug 10 '21 at 13:15