The Triangular Decomposition $h\oplus n_+\oplus n_-$ of $M_2(\mathbb{C})$
Introduction
The algebra splits into the diagonal algebra of dimension $2$ and the two strictly triangular lines, and the brackets of the three pieces form the root system $A_{1}$.
The article proves the decomposition $M_2(\mathbb{C})=h\oplus n_+\oplus n_-$, identifies $h\oplus n_+$ with the algebra $UT_2$ of upper triangular matrices and with the stabiliser of a line, computes the brackets between the three pieces, reads the roots $\pm\alpha$ and the coroot, and shows that $n_+$ is the nilradical of $UT_2$. It assumes The Fine Decomposition against the Diagonal at $n=2$ for the corner coordinates and The Ideals, the Centre and the Simplicity of $M_2(\mathbb{C})$ for the ideals.
It defers the antisymmetric part inside $UT_2$ to The Antisymmetric Part inside $UT_2$: the Radical as the Commutator and the warning about the wrong product to $UT_2$ as a Linear Model of $\operatorname{Sym}_2$: Same Space, Wrong Product. The biquaternion version of the same split is The Peirce Basis of the Biquaternion Algebra, where the two triangular lines are the mixed corners.
The Decomposition
Definition. In the standard frame the three pieces are
$$ h=\left\{\begin{pmatrix}a&0\\ 0&d\end{pmatrix}\right\},\qquad n_+=\mathbb{C}E_{12}=\left\{\begin{pmatrix}0&b\\ 0&0\end{pmatrix}\right\},\qquad n_-=\mathbb{C}E_{21}=\left\{\begin{pmatrix}0&0\\ c&0\end{pmatrix}\right\}. $$
Theorem. The sum is direct and equal to the algebra:
$$ M_2(\mathbb{C})=h\oplus n_+\oplus n_-,\qquad \dim_{\mathbb{C}}=2+1+1=4 . $$
Proof. Every matrix is $\begin{pmatrix}a&b\\ c&d\end{pmatrix}$, with $a,d$ in $h$ and $b,c$ in the two lines; independence follows by reading the four entries separately.
Corollary. $M_2(\mathbb{C})=(h\oplus n_+)\oplus n_-$ is also the sum of the upper triangular algebra and the strictly lower line, and that is the form used in $UT_2$ as the Complement of the Antisymmetric Part.
The decomposition refines the split into the two halves. The diagonal $h$ lies inside the symmetric half, and each of the two lines $n_\pm$ splits into a symmetric and an antisymmetric part.
The Algebra $UT_2$
Definition. $UT_2=h\oplus n_+$ is the algebra of upper triangular matrices.
Proposition. $UT_2$ is a subalgebra of dimension $3$, the line $n_+$ is a two-sided ideal of it, and $UT_2/n_+\cong h\cong\mathbb{C}\times\mathbb{C}$.
Proof. The product of two upper triangular matrices is upper triangular, and the product of an upper triangular matrix with a strictly upper one is strictly upper. The quotient by $n_+$ is the image of the diagonal, and diagonal matrices multiply entrywise.
Proposition. $UT_2=\{A:A(\mathbb{C}e_{1})\subseteq\mathbb{C}e_{1}\}$ is the stabiliser of the line $\mathbb{C}e_{1}$, so it is the parabolic subalgebra attached to that line.
Proof. The first column of $A$ is the image of $e_{1}$, and the condition that it be a multiple of $e_{1}$ is $c=0$, that is $A\in h\oplus n_+$.
Corollary. The parabolic subalgebras of $M_2(\mathbb{C})$ are the stabilisers of the lines of $\mathbb{C}^{2}$; they are parametrised by $\mathbb{P}^{1}$ and they are all conjugate.
Proof. Each line gives one stabiliser, and $\mathrm{GL}_2(\mathbb{C})$ acts transitively on the lines.
The nilradical. The line $n_+$ is nilpotent, since $E_{12}^{2}=0$, and it is the largest nilpotent ideal of $UT_2$; so the nilpotent radical of $UT_2$ is $n_+$ and $UT_2$ is not reductive.
The Brackets of the Three Pieces
Theorem. The three pieces satisfy
$$ [h,h]=0,\qquad [h,n_+]\subseteq n_+,\qquad [h,n_-]\subseteq n_-,\qquad [n_+,n_-]\subseteq h,\qquad [n_+,n_+]=[n_-,n_-]=0 . $$
Proof. The diagonal matrices commute and act diagonally on the two matrix units: for $D=\operatorname{diag}(a,d)$ one has $[D,E_{12}]=(a-d)E_{12}$ and $[D,E_{21}]=(d-a)E_{21}$. The bracket $[E_{12},E_{21}]=E_{11}-E_{22}$ is diagonal, and a one-dimensional space brackets to zero with itself.
Corollary. The traceless diagonal $h_{0}=\mathbb{C}(E_{11}-E_{22})$ is a Cartan subalgebra of $\mathfrak{sl}_2(\mathbb{C})$, and $\mathfrak{sl}_2(\mathbb{C})=h_{0}\oplus n_+\oplus n_-$ is its root space decomposition.
Proof. The three spaces lie inside the traceless algebra, their dimensions are $1+1+1=3$, and the brackets above are the root space axioms.
The roots are $\pm\alpha$. Define $\alpha\in h_{0}^{*}$ by $\alpha(\operatorname{diag}(a,-a))=2a$. Then $[X,E_{12}]=\alpha(X)E_{12}$ and $[X,E_{21}]=-\alpha(X)E_{21}$ for $X\in h_{0}$, so the two lines are the root spaces of $\alpha$ and of $-\alpha$.
The coroot. With $H=E_{11}-E_{22}$ one has $\alpha(H)=2$ and $\operatorname{tr}(H^{2})=2$, so $\langle H,X\rangle=\alpha(X)$ for every $X\in h_{0}$; the element $H$ is the coroot of $\alpha$, and the normalisation by the trace form is what makes the pairing return the structural constant $2$.
The root system is $A_{1}$. There are two roots, opposite to one another, and the Weyl group is the group of order two that exchanges them. At order $n$ the root system is $A_{n-1}$ and the Weyl group is the symmetric group, which is The Matrix Algebra $M_n(\mathbb{C})$.
The Triangular Decomposition and the Flag
Proposition. The decomposition records the flag $0\subset\mathbb{C}e_{1}\subset\mathbb{C}^{2}$: the line $n_+$ moves the line, the diagonal $h$ preserves it, and the line $n_-$ sends it to $0$.
Proof. $E_{12}$ sends $e_{1}$ to $e_{2}$, which leaves the line; a diagonal matrix keeps $e_{1}$ inside the line; $E_{21}$ kills $e_{1}$.
Corollary. $n_+\oplus n_-$ is the space of matrices with zero diagonal, of dimension $2$, that is the tangent space to the set of matrices killing the flag.
Proof. A matrix of $n_+\oplus n_-$ has both diagonal entries zero, and conversely a matrix with zero diagonal lies in the sum of the two lines.
The split is orthogonal in one direction only. The diagonal $h$ is orthogonal to the off-diagonal plane for the trace form, since $\operatorname{tr}(E_{ii}E_{12})=\operatorname{tr}(E_{ii}E_{21})=0$; but $n_+$ and $n_-$ are not orthogonal, since $\operatorname{tr}(E_{12}E_{21})=1$.
Proof. The two computations are immediate from $\operatorname{tr}(E_{ij}E_{kl})=\delta_{jk}\delta_{il}$.
The Two Lines and the Two Halves
Proposition. The symmetric half meets $n_+$ in the real line spanned by $E_{12}+E_{21}$ and meets $n_-$ in the same real line; the antisymmetric half meets $n_+$ in the real line spanned by $E_{12}-E_{21}$.
Proof. A matrix $bE_{12}+cE_{21}$ is symmetric exactly when $b=c$, and antisymmetric exactly when $b=-c$.
Corollary. Inside the off-diagonal plane the two halves are two lines, orthogonal for the trace form and exchanged by a rotation of angle $\pi/4$ of that plane.
Proof. The pairing of $E_{12}+E_{21}$ with $E_{12}-E_{21}$ is $\operatorname{tr}(E_{12}E_{21}-E_{21}E_{12})=1-1=0$.
Corollary. $UT_2$ contains the two-dimensional part $h\oplus\mathbb{R}(E_{12}+E_{21})$ of the symmetric half, and its third direction $E_{12}-E_{21}$ is antisymmetric.
Proof. The diagonal is symmetric and $E_{12}+E_{21}$ is symmetric, while $E_{12}-E_{21}$ is antisymmetric and completes a basis with them.
The antisymmetric direction is the radical. In $UT_2$ the antisymmetric line is exactly the commutator subspace, which is the content of The Antisymmetric Part inside $UT_2$: the Radical as the Commutator.
Summary
$M_2(\mathbb{C})$ splits as the diagonal algebra of dimension $2$ and the two strictly triangular lines $n_+=\mathbb{C}E_{12}$ and $n_-=\mathbb{C}E_{21}$. The sum $h\oplus n_+$ is the upper triangular algebra $UT_2$, of dimension $3$, the stabiliser of the line $\mathbb{C}e_{1}$ and the parabolic attached to it, with nilradical $n_+$.
The brackets are $[h,n_\pm]\subseteq n_\pm$ and $[n_+,n_-]\subseteq h$, and each of the two nilpotent lines brackets to zero with itself. The traceless diagonal is a Cartan subalgebra of $\mathfrak{sl}_2(\mathbb{C})$ with roots $\pm\alpha$, where $\alpha(\operatorname{diag}(a,-a))=2a$, coroot $E_{11}-E_{22}$, and root system $A_{1}$. The diagonal is orthogonal to the off-diagonal plane, but the two nilpotent lines pair to $1$ and are not orthogonal. Inside the off-diagonal plane the symmetric and the antisymmetric halves are two orthogonal lines.
Summary of Notation
| Symbol | Meaning |
|---|---|
| $h$ | the diagonal algebra, dimension $2$ |
| $n_+$ | $\mathbb{C}E_{12}$, the strictly upper line |
| $n_-$ | $\mathbb{C}E_{21}$, the strictly lower line |
| $UT_2$ | $h\oplus n_+$, the upper triangular algebra |
| $h_{0}$ | the traceless diagonal, a Cartan subalgebra |
| $\alpha$ | the root, $\alpha(\operatorname{diag}(a,-a))=2a$ |
| $H$ | $E_{11}-E_{22}$, the coroot direction |
| $A_{1}$ | the root system of $\mathfrak{sl}_2$ |
Further Reading
- James E. Humphreys, Introduction to Lie Algebras and Representation Theory (Springer, 1972), for root space decompositions and the system $A_{1}$.
- Armand Borel, Linear Algebraic Groups (Springer, 1991), for parabolics as stabilisers of flags.
- The Fine Decomposition against the Diagonal at $n=2$, for the corner coordinates used here.
- The Two Halves over $\mathbb{R}$: Symmetric and Antisymmetric at $n=2$, for the symmetric and antisymmetric lines inside the triangular pieces.
- The Peirce Basis of the Biquaternion Algebra, for the mixed corners in the biquaternion language.