The Spin-One Carrier of $M_n(\mathbb{C})$: the Three-Dimensional $V_1$

Introduction

The question. The decomposition $M_n=\bigoplus_{k=0}^{n-1}V_k$ has a summand of dimension three for every $n\ge2$. What is it, where does it sit, and what does it generate?

The answer in one line. $V_1$ is the span of the three generators $J_x,J_y,J_z$, it is the adjoint copy of the principal $\mathfrak{su}_2$ inside $\mathfrak{sl}_n$, and its multiplicative closure is the whole algebra.

Why the article is needed. The carrier $V_1$ is the only carrier that can be written down without a projector: it is the span of the three matrices that define the principal copy in the first place. That makes it the natural first check of the decomposition, and the natural place where the biquaternion "vector part" is compared with the general case.

Boundaries. The decomposition is The Spin Decomposition of $M_n(\mathbb{C})$; the carrier of spin zero is The Spin-Zero Carrier of $M_n(\mathbb{C})$; the algebra generated by the generators is The Carrier $V_1$ Generates the Algebra, a section of the decomposition article; the biquaternion vector part is The Spin-One Anti-Hermitian Subspace.

Conventions. $J_x,J_y,J_z$ are the images in $\mathfrak{gl}_n$ of the standard generators of $\mathfrak{su}_2$ under $\operatorname{Sym}^{n-1}$, taken Hermitian, with $[J_i,J_j]=\sum_k\varepsilon_{ijk}J_k$.

The Subspace

Proposition. $\operatorname{span}\{J_x,J_y,J_z\}$ is a three-dimensional subspace of $\mathfrak{sl}_n$, invariant under the adjoint action, hence equal to $V_1$.

Proof. The three matrices are linearly independent, since their squares are distinct linear combinations and no nontrivial combination vanishes on the weight basis. They are traceless, and the bracket of two of them is a combination of the three, so the span is invariant. The carrier $V_1$ is the unique three-dimensional invariant subspace of $M_n$, because the decomposition is multiplicity-free. $\square$

The dimension. $\dim V_1=3$ at every $n\ge2$.

The Casimir. $-\sum_i\operatorname{ad}_{J_i}^2$ (the $J_i$ anti-Hermitian) acts on $V_1$ as $1\cdot2=2$, the smallest nonzero value.

The weights. The weights of $V_1$ are $-1,0,1$, one each.

The bracket. $[V_1,V_1]\subseteq V_1$, because $V_1$ is a Lie subalgebra: it is the copy of $\mathfrak{su}_2$ itself.

The reading. $V_1$ is the adjoint module. The algebra acts on it by the bracket, and the bracket does not leave it, which is the statement that $\mathfrak{su}_2$ is a Lie algebra and not merely a subspace.

Where It Sits

Inside $\mathfrak{sl}_n$. $V_1\subset\mathfrak{sl}_n$, because the generators are traceless. The traceless part is $\bigoplus_{k=1}^{n-1}V_k$, and $V_1$ is its lowest summand.

Not the centre. $V_1\cap V_0=0$, since the identity is not traceless.

The orthogonal complement. $V_1$ is orthogonal to $V_k$ for $k\neq1$ for the Hilbert–Schmidt form, because the carriers are inequivalent and the form is invariant.

The comparison at $n=2$. $\mathfrak{sl}_2$ has dimension three and equals $V_1$: at $n=2$ the adjoint copy is the whole traceless part. At $n\ge3$ the traceless part is larger, and $V_1$ is a proper subspace of it.

The consequence. The statement "$V_1=\mathfrak{sl}_n$" is true only at $n=2$. It is the second biquaternion coincidence, after the coincidence of labels, and it fails at $n=3$ where $\mathfrak{sl}_3=V_2\oplus V_1$ has dimension eight.

What It Generates

Proposition. The algebra generated by $V_1$ under multiplication is the whole of $M_n$.

Proof. $V_1$ contains the three generators, and the three generators generate $M_n$ as an algebra by Burnside's theorem. Any algebra containing $V_1$ contains the generators, hence is $M_n$. $\square$

The dimension table. The closure of $V_1$ has dimension $n^2$: $4,9,16,25$ for $n=2,3,4,5$.

The contrast with $V_0$. The closure of $V_0$ is $V_0$ itself, of dimension one. The two extreme carriers behave in opposite ways: the smallest closes to a point, the next closes to everything.

The intermediate carriers. The closures of $V_k$ for $k\ge2$ are also $M_n$, so every nonzero carrier except $V_0$ generates the algebra.

The Biquaternion Vector Part

The real case. The pure imaginary quaternions form a real three-dimensional subspace of $\mathbb{B}$, closed under the bracket, and they are the adjoint copy of the rotor group.

The complexification. $\mathbb{B}\otimes\mathbb{C}\cong M_2(\mathbb{C})$, and the complexified vector part is a complex three-dimensional subspace of $M_2$, invariant and irreducible: it is $V_1$.

What is the same. Dimension three, the adjoint action, the bracket, and the closure equal to the whole algebra.

What is different. The biquaternion vector part is a real subspace of a real algebra; $V_1$ is a complex subspace of a complex algebra. At $n=2$ they are the same object read over the two fields. At $n\ge3$ the complex $V_1$ has no real form inside a real algebra of the same shape, because there is no real algebra whose complexification is $M_n$ and whose dimension is $n^2$ for $n\ge3$.

The honest statement. The vector part of the biquaternion algebra is the case $n=2$ of $V_1$; the "real three-dimensional" description is a property of that case, not of the carrier.

Is $V_1$ Unique in Its Role?

The uniqueness of the carrier. $V_1$ is the only carrier of dimension three, because the carriers have dimensions $1,3,5,\dots,2n-1$ and are distinct.

The non-uniqueness of the role. Being the adjoint module is a role that $V_1$ plays for the principal copy. For another group the adjoint module is another carrier: under the block reading at $n=4$ the adjoint of $SU(2)\times SU(2)$ involves the $(1,0)$ and $(0,1)$ summands, and under the full $U(n)$ the adjoint is the whole traceless part.

The consequence. "The spin-one carrier" is a name that depends on the group, exactly as "the spin content" does.

The Finite Check

Checked for $n=2,3,4,5$: the three generators are linearly independent and traceless; the span is invariant under all three adjoint maps, with residual below $10^{-14}$; the Casimir acts on the span as the scalar $2$ with residual below $10^{-13}$.

The weight multiset of the span is $\{-1,0,1\}$ at every $n$ from $2$ to $5$, computed from $\operatorname{ad}_{J_z}$.

The closure was computed by iterating products of a basis: the algebra generated by $V_1$ has dimension $4,9,16,25$, equal to $n^2$, while the algebra generated by $V_0$ has dimension $1$.

The Real Form Inside $V_1$

The generators. The $iJ_x,iJ_y,iJ_z$ are anti-Hermitian and their real span is closed under the bracket, with $[iJ_x,iJ_y]=-iJ_z$.

The isomorphism. That real span is a real form of $V_1$ isomorphic to $\mathfrak{so}(3)$, and $V_1=\mathfrak{so}(3)\otimes\mathbb{C}$.

The reading. The spin-one carrier is the complexification of the adjoint Lie algebra, and the Hermitian generators $J_i$ span the complementary real subspace, not a real form.

The Adjoint Representation

The statement. The copy acts on $V_1$ by the adjoint representation, which is the three-dimensional irreducible representation with kernel $\{\pm\mathrm I\}$ at every $n$.

The contrast with the module. The module is faithful for even $n$ and has kernel $\{\pm\mathrm I\}$ for odd $n$; the adjoint carrier always has the kernel.

The reading. $V_1$ is the carrier on which the copy always acts through $SO(3)$, whatever the parity of $n$.

The Products

The statement. The product of two elements of $V_1$ lies in $V_0\oplus V_1\oplus V_2$.

The check. The residual of $V_1\cdot V_1$ against $V_0\oplus V_1\oplus V_2$ is zero at $n=2,3,4,5$, computed by projecting the products of a basis onto the Casimir eigenspaces.

The reading. The truncation of the Clebsch–Gordan rule at $1\otimes1=0\oplus1\oplus2$ is exact for the carrier $V_1$, and the same truncation is what makes the product of two elements of $V_k$ land in the carriers up to $V_{2k}$.

The consequence. The bracket $[V_1,V_1]\subseteq V_1$ is the antisymmetric part of the product, and the symmetric part is the sum of the scalar and the spin-two pieces.

The Commutator Table

The brackets. $[J_x,J_y]=iJ_z$ and its cyclic permutations, with the Hermitian generators.

The real form. The anti-Hermitian multiples $iJ_i$ close under the bracket with real structure constants, and they are the real form $\mathfrak{so}(3)$.

The reading. The complex carrier and its real form differ by the factor $i$, and the factor is the whole difference between the Lie algebra over $\mathbb{C}$ and over $\mathbb{R}$.

The Two Extreme Carriers

The smallest. $V_0$: dimension one, Casimir zero, closure one, no basis to choose.

The next. $V_1$: dimension three, Casimir two, closure $M_n$, spanned by the three generators.

The reading. The two smallest carriers are the two extremes of behaviour, and every carrier above $V_1$ behaves like $V_1$ rather than like $V_0$.

The Trace

The statement. Every element of $V_1$ is traceless, and the trace pairing with $V_0$ vanishes.

The consequence. $V_1$ is orthogonal to the scalars for the Hilbert–Schmidt form, so the carrier sits inside $\mathfrak{sl}_n$ and not merely inside $M_n$.

The reading. The orthogonality is the reason the decomposition of the algebra splits cleanly into the scalars and the traceless part, with $V_1$ as the first traceless carrier.

Summary

$V_1$ is the span of the three generators, the adjoint copy of the principal $\mathfrak{su}_2$ inside $\mathfrak{sl}_n$, of dimension three and Casimir value two. Its bracket stays inside it and its multiplicative closure is $M_n$. At $n=2$ it is the whole traceless part; at $n\ge3$ it is a proper subspace. The biquaternion vector part is the real form of the same object at $n=2$.

Summary of Notation

Symbol Meaning
$V_1$ $\operatorname{span}\{J_x,J_y,J_z\}$
$k(k+1)=2$ the Casimir value on $V_1$
$[V_1,V_1]\subseteq V_1$ $V_1$ is a Lie subalgebra
$\overline{V_1}=M_n$ the multiplicative closure

Further Reading