The Element, the Coordinates and the Trace of $M_5(\mathbb{C})$

Introduction

An element of $M_5(\mathbb{C})$ is a matrix $A=(A_{ij})$, and the twenty-five entries are its coordinates in the basis of matrix units:

$$ A=\sum_{i,j=1}^{5}A_{ij}E_{ij}. $$

The coordinates are complex numbers; separating real and imaginary parts gives fifty real coordinates. The trace is the linear form

$$ \operatorname{tr}A=\sum_{i=1}^{5}A_{ii}, $$

the sum of the diagonal coordinates. It is the unique (up to scale) linear form invariant under conjugation, and it defines the trace pairing $\langle A,B\rangle=\operatorname{tr}(AB)$, a symmetric nondegenerate complex bilinear form.

This article develops the element from the coordinate side: the matrix units and the coordinates, the real and complex readings, the trace as an invariant linear form, the trace pairing and its dual, the hyperplane $\mathfrak{sl}_5$ of traceless matrices, and the Jordan decomposition $A=S+N$ with its trace. The trace pairing is not the positive definite form of the Hilbert space; that form is $\operatorname{tr}(A^*B)$, defined with the conjugate transpose in The Involution and the Hermitian Structure of $M_5(\mathbb{C})$. The present article uses the complex bilinear pairing of the algebra, the one that pairs $E_{ij}$ with $E_{ji}$.

The article assumes the matrix algebra of The Matrix Algebra $M_5(\mathbb{C})$ and the root of Roots of Unity and the Lagrange Idempotents of $M_5(\mathbb{C})$. The Jordan decomposition is stated and proved in the finite-dimensional case; the spectral theorem for the Hermitian form, the polar decomposition and the exponential are later articles.

The Coordinates

The Matrix Units

Definition. The coordinates of $A$ are the numbers $A_{ij}$ in the expansion

$$ A=\sum_{i,j=1}^{5}A_{ij}E_{ij}. $$

The coordinate functional $x_{ij}:M_5\to\mathbb{C}$ is $x_{ij}(A)=A_{ij}$; it is the dual basis element to $E_{ij}$.

Proposition. The map $A\mapsto(A_{ij})$ is a linear isomorphism $M_5(\mathbb{C})\cong\mathbb{C}^{25}$. The coordinate functionals satisfy $x_{ij}(E_{kl})=\delta_{ik}\delta_{jl}$, and the product is

$$ (AB)_{ij}=\sum_{k=1}^{5}A_{ik}B_{kj}. $$

Proof. The matrix units are a basis, so the expansion is unique; the coordinate of $E_{kl}$ is the Kronecker delta; the product formula is the definition of the matrix product.

Corollary. The algebra is a complex vector space of dimension $25$ and a real vector space of dimension $50$. A real basis is $E_{ij}$ and $iE_{ij}$ for $1\le i,j\le5$.

Proof. The $25$ complex coordinates give $25$ complex dimensions; separating real and imaginary parts gives $50$ real coordinates. The real basis is the $25$ units and their $i$-multiples.

The Real and Complex Readings

Definition. Write $A_{ij}=x_{ij}+iy_{ij}$ with $x_{ij},y_{ij}\in\mathbb{R}$. The real coordinates of $A$ are the fifty numbers $(x_{ij},y_{ij})$.

Proposition. The multiplication is polynomial in the real coordinates, and the conjugation $A\mapsto\bar A$ is the real-linear map $(x_{ij},y_{ij})\mapsto(x_{ij},-y_{ij})$. The algebra is a real algebra of dimension $50$ whose complexification is itself.

Proof. The product formula is polynomial with real coefficients in the real and imaginary parts; the conjugation negates the imaginary part. The complex structure is the multiplication by $i$, a real-linear map $J$ with $J^2=-1$; the real algebra with $J$ is the complex algebra.

Remark. The real dimension $50$ is the dimension of the algebra as a real vector space, not the dimension of a real form. The real form $\mathfrak{u}(5)$ has dimension $25$ and is not the whole algebra; it is the anti-Hermitian half of The Matrix Algebra. The fifty real coordinates are the coordinates of the complex algebra read over $\mathbb{R}$; the twenty-five complex coordinates are the coordinates over $\mathbb{C}$.

The Element as a Vector

Definition. The element $A$ is the vector of its coordinates; the scalar part is $\frac15\operatorname{tr}(A)I$, and the traceless part is $A-\frac15\operatorname{tr}(A)I$.

Proposition. The map

$$ A\longmapsto\Bigl(\frac15\operatorname{tr}(A)I,\ A-\frac15\operatorname{tr}(A)I\Bigr) $$

is a direct-sum decomposition

$$ M_5(\mathbb{C})=\mathbb{C}I\oplus\mathfrak{sl}_5(\mathbb{C}), \qquad \mathfrak{sl}_5=\{A:\operatorname{tr}A=0\}, $$

of complex dimensions $1$ and $24$.

Proof. The map is linear; the scalar part is a multiple of $I$ and the traceless part has trace $\operatorname{tr}A-\frac15\operatorname{tr}(A)\operatorname{tr}I=0$. The intersection is zero: a scalar matrix $\lambda I$ with trace $5\lambda=0$ has $\lambda=0$. The dimensions are $1$ and $25-1=24$.

Corollary. The trace is the coordinate of the scalar part: $\operatorname{tr}A=5\lambda$ for the scalar part $\lambda I$. The hyperplane $\mathfrak{sl}_5$ is the kernel of the trace.

Proof. The trace of $\lambda I$ is $5\lambda$; the traceless part is in the kernel by definition.

The Trace

The Invariant Linear Form

Theorem. The linear forms $\ell:M_5(\mathbb{C})\to\mathbb{C}$ invariant under conjugation,

$$ \ell(gAg^{-1})=\ell(A)\qquad\text{for all }g\in GL_5(\mathbb{C}), $$

form the one-dimensional space $\mathbb{C}\operatorname{tr}$.

Proof. Let $\ell$ be invariant. Differentiating the identity $\ell(gAg^{-1})=\ell(A)$ at $g=I$ in the direction $X$ gives

$$ \ell([X,A])=0\qquad\text{for all }X,A . $$

The commutators span the traceless subspace: for $i\ne j$, $E_{ij}=[E_{ii},E_{ij}]$, and $E_{ii}-E_{jj}=[E_{ij},E_{ji}]$; these elements span $\mathfrak{sl}_5$. Hence $\ell$ vanishes on $\mathfrak{sl}_5$, and $\ell(A)=\ell(\frac15\operatorname{tr}(A)I)=\frac15\operatorname{tr}(A)\ell(I)$. So $\ell$ is a multiple of the trace. Conversely the trace is invariant, $\operatorname{tr}(gAg^{-1})=\operatorname{tr}(A)$.

Corollary. The trace is the unique invariant linear form with $\operatorname{tr}I=5$. It is the character of the defining representation of $GL_5$.

Proof. The normalization fixes the multiple; the trace of the identity is $5$.

The Trace Pairing

Definition. The trace pairing is

$$ \langle A,B\rangle=\operatorname{tr}(AB). $$

Proposition. The trace pairing is complex bilinear, symmetric and nondegenerate. In the matrix-unit basis,

$$ \langle E_{ij},E_{kl}\rangle=\delta_{il}\delta_{jk}, $$

so the dual basis of $E_{ij}$ is $E_{ji}$.

Proof. Bilinearity is the linearity of the trace and of the product; symmetry is $\operatorname{tr}(AB)=\operatorname{tr}(BA)$; the matrix-unit value is $\operatorname{tr}(E_{ij}E_{kl})=\delta_{jk}\operatorname{tr}(E_{il})=\delta_{jk}\delta_{il}$. If $\langle A,B\rangle=0$ for all $B$, testing $B=E_{ji}$ gives $A_{ij}=0$; hence $A=0$.

Corollary. The trace pairing identifies $M_5$ with its dual by $A\mapsto\langle A,\cdot\rangle$; the inverse identification sends a linear form $\ell$ to the matrix $\sum_{i,j}\ell(E_{ij})E_{ji}$.

Proof. The pairing is nondegenerate, so the map to the dual is an isomorphism; the inverse is read on the basis.

The Orthogonality and the Adjoint

Definition. The orthogonal complement of a subspace $W$ is

$$ W^{\perp}=\{A:\langle A,B\rangle=0\text{ for all }B\in W\}. $$

Proposition. $\dim W+\dim W^{\perp}=25$, and $(W^{\perp})^{\perp}=W$.

Proof. The pairing is nondegenerate; the orthogonal complement is the annihilator of $W$ under the isomorphism with the dual, of dimension $25-\dim W$; the double complement is $W$.

Proposition. The trace pairing is invariant under conjugation and under the adjoint action:

$$ \langle gAg^{-1},gBg^{-1}\rangle=\langle A,B\rangle,\qquad \langle [A,B],C\rangle+\langle B,[A,C]\rangle=0 . $$

Proof. The first identity is the cyclic invariance of the trace: $\operatorname{tr}(gAg^{-1}gBg^{-1})=\operatorname{tr}(gABg^{-1})=\operatorname{tr}(AB)$. For the second, $\langle [A,B],C\rangle=\operatorname{tr}(ABC-BAC)$ and $\langle B,[A,C]\rangle=\operatorname{tr}(BAC-BCA)$; the sum is $\operatorname{tr}(ABC-BCA)=0$ by cyclicity.

Corollary. The adjoint operator $\operatorname{ad}_A$ is anti-self-adjoint for the trace pairing: $\langle\operatorname{ad}_A B,C\rangle=-\langle B,\operatorname{ad}_A C\rangle$. The trace form is the invariant bilinear form of the Lie algebra $\mathfrak{gl}_5$.

Proof. The second identity is exactly the anti-self-adjointness.

Remark. The trace pairing is indefinite. It is a complex bilinear form, not a Hermitian form; $\langle A,A\rangle=\operatorname{tr}(A^2)$ can be zero for a nonzero nilpotent $A$. The positive definite form is $\operatorname{tr}(A^*B)$, which uses the conjugate transpose and is treated in The Involution and the Hermitian Structure of $M_5(\mathbb{C})$. The trace pairing is the algebraic pairing of the Frobenius algebra; the definite form is the metric of the Hilbert space.

The Jordan Decomposition

The Semisimple and Nilpotent Parts

Definition. An element $S$ is semisimple if it is diagonalizable; an element $N$ is nilpotent if $N^k=0$ for some $k$.

Theorem (Jordan decomposition). Every $A\in M_5(\mathbb{C})$ has a unique decomposition

$$ A=S+N,\qquad S\text{ semisimple},\quad N\text{ nilpotent},\quad SN=NS . $$

Moreover $S=p(A)$ for a polynomial $p$, and $N=A-S$.

Proof. Let $\chi_A(t)=\prod_{k=1}^{r}(t-\lambda_k)^{m_k}$ be the characteristic polynomial with distinct roots $\lambda_k$. By the Chinese remainder theorem, choose a polynomial $p$ with

$$ p(t)\equiv\lambda_k\pmod{(t-\lambda_k)^{m_k}},\qquad k=1,\dots,r . $$

Then $S=p(A)$ commutes with $A$, and $S$ acts on the generalized eigenspace of $\lambda_k$ as $\lambda_k$; hence $S$ is semisimple with the same eigenvalues. The difference $N=A-S$ is nilpotent on each generalized eigenspace, because $A-\lambda_kI$ is nilpotent there and $S-\lambda_kI$ vanishes there; hence $N$ is nilpotent. The parts commute because $S$ is a polynomial in $A$. For uniqueness, if $A=S'+N'$ with $S'$ semisimple, $N'$ nilpotent and $S'N'=N'S'$, then $S'$ commutes with $A$ and preserves each generalized eigenspace; on the generalized eigenspace of $\lambda_k$, $S'$ has the single eigenvalue $\lambda_k$ (because $N'$ is nilpotent and commutes with $S'$), so $S'=S$ and $N'=N$.

Corollary. The trace of $A$ is the sum of its eigenvalues with multiplicity, and the trace of $N$ is zero. The determinant is the product of the eigenvalues with multiplicity.

Proof. The trace is the sum of the diagonal entries in a triangular form; the eigenvalues are the diagonal entries with multiplicity; $N$ is strictly triangular in a suitable basis, so its trace is zero. The determinant is the product of the diagonal entries.

The Jordan Form

Theorem. Every $A$ is similar to a Jordan matrix: a direct sum of Jordan blocks $J_m(\lambda)$, of size $m$ and eigenvalue $\lambda$, with $1$ on the superdiagonal and $\lambda$ on the diagonal. The semisimple part $S$ is the diagonal part and the nilpotent part $N$ the superdiagonal part of the Jordan form.

Proof. The standard Jordan form theorem; the generalized eigenspaces decompose into cyclic subspaces for $A-\lambda I$, each giving a Jordan block. The diagonal part is semisimple, the superdiagonal part nilpotent, and they commute.

Corollary. The trace depends only on the eigenvalues and their multiplicities; the nilpotent part does not contribute. Two matrices with the same characteristic polynomial have the same trace and determinant, but not necessarily the same Jordan form.

Proof. The trace is the sum of the diagonal entries, which are the eigenvalues; the Jordan form exhibits the diagonal part. The characteristic polynomial records the eigenvalues with multiplicity, not the block sizes.

The Trace and the Decomposition

Proposition. For the Jordan decomposition $A=S+N$,

$$ \operatorname{tr}A=\operatorname{tr}S,\qquad \operatorname{tr}N=0,\qquad \det A=\det S . $$

Proof. $N$ is nilpotent, hence all its eigenvalues are zero, hence its trace is zero and its determinant is zero; the determinant of $A$ is the product of the eigenvalues, which are those of $S$ with multiplicity. The trace of $A$ is the trace of $S$ plus the trace of $N$.

Remark. The trace is blind to the nilpotent part. The nilpotent part is visible in the Jordan form and in the minimal polynomial, not in the trace or the determinant. The characteristic polynomial records the eigenvalues with multiplicity; the trace and the determinant are the first and last coefficients. The intermediate coefficients are the elementary symmetric functions, which also see only the eigenvalues.

The Trace and the Determinant

The Differential

Proposition. The differential of the determinant at the identity is the trace:

$$ \det(I+tA)=1+t\operatorname{tr}A+O(t^2),\qquad d\det_I(A)=\operatorname{tr}A . $$

Proof. Expand the determinant of $I+tA$ by the Leibniz formula; the identity permutation contributes $1$; the transpositions contribute terms of order $t^2$ or higher; the only order-$t$ term is the product of the diagonal entries $1+tA_{ii}$, whose sum is $1+t\operatorname{tr}A$. The higher-order terms are $O(t^2)$.

Corollary. The determinant is a local coordinate near the identity, and its level set $\det=1$ is a hypersurface of dimension $24$ through $I$; the tangent space is $\mathfrak{sl}_5$.

Proof. The derivative is the nonzero linear form $\operatorname{tr}$; its kernel is $\mathfrak{sl}_5$, of dimension $24$; the level set is a smooth hypersurface by the implicit function theorem.

The Exponential Identity

Theorem. For every $A\in M_5(\mathbb{C})$,

$$ \det(e^A)=e^{\operatorname{tr}A}. $$

Proof. Put $A$ in Jordan form, $A=gJg^{-1}$. Then $e^A=ge^Jg^{-1}$, and both sides are invariant under conjugation: $\det(e^A)=\det(e^J)$ and $\operatorname{tr}A=\operatorname{tr}J$. For a Jordan block $J_m(\lambda)=\lambda I+N$ with $N$ nilpotent, $e^{J_m(\lambda)}=e^\lambda e^N$, and $e^N$ is unipotent with determinant $1$; hence $\det(e^{J_m(\lambda)})=e^{m\lambda}$. The product over the blocks gives $e^{\operatorname{tr}J}=e^{\operatorname{tr}A}$.

Corollary. The exponential maps the traceless hyperplane $\mathfrak{sl}_5$ into the special linear group $SL_5$, and $\det(e^A)=1$ if and only if $\operatorname{tr}A\in2\pi i\mathbb{Z}$.

Proof. $\det(e^A)=e^{\operatorname{tr}A}=1$ iff $\operatorname{tr}A\in2\pi i\mathbb{Z}$; in particular a traceless $A$ gives determinant $1$.

Summary

An element $A$ has $25$ complex coordinates $A_{ij}$ in the matrix-unit basis, or $50$ real coordinates. The trace is the sum of the diagonal coordinates; it is the unique conjugation-invariant linear form up to scale, and it defines the symmetric nondegenerate complex bilinear pairing $\langle A,B\rangle=\operatorname{tr}(AB)$, with dual basis $E_{ji}$ to $E_{ij}$.

$$ \boxed{\ \dim_{\mathbb{C}}M_5=25,\quad \dim_{\mathbb{R}}M_5=50,\quad \operatorname{tr}\text{ is the unique invariant linear form},\quad \langle E_{ij},E_{kl}\rangle=\delta_{il}\delta_{jk}.\ } $$

The algebra splits as $\mathbb{C}I\oplus\mathfrak{sl}_5$, of dimensions $1+24$. Every element has the Jordan decomposition $A=S+N$ with $S$ semisimple, $N$ nilpotent and $SN=NS$; the trace is $\operatorname{tr}S$, the nilpotent part has trace zero, and $\det A=\det S$. The differential of the determinant at $I$ is the trace, and $\det(e^A)=e^{\operatorname{tr}A}$.

Object Value
Complex coordinates $25$
Real coordinates $50$
Trace $\sum_iA_{ii}$
Invariant linear forms $\mathbb{C}\operatorname{tr}$
Trace pairing $\operatorname{tr}(AB)$, symmetric, nondegenerate, indefinite
Dual of $E_{ij}$ $E_{ji}$
Traceless hyperplane $\mathfrak{sl}_5$ $24$
Jordan decomposition $A=S+N$, $SN=NS$
Trace of $N$ $0$
Differential of $\det$ at $I$ $\operatorname{tr}$
Exponential identity $\det(e^A)=e^{\operatorname{tr}A}$

Summary of Notation

Symbol Meaning
$A_{ij}$ The coordinates of $A$ in the matrix-unit basis
$x_{ij}$ The coordinate functional $A\mapsto A_{ij}$
$\operatorname{tr}A$ The trace, $\sum_iA_{ii}$
$\langle A,B\rangle=\operatorname{tr}(AB)$ The trace pairing
$\mathfrak{sl}_5$ The traceless matrices, dimension $24$
$S,N$ The semisimple and nilpotent parts of $A$
$J_m(\lambda)$ The Jordan block of size $m$ and eigenvalue $\lambda$

Further Reading