Spherical Tensor Operators

States organize under rotations into $|j,m\rangle$ multiplets — that was the spine of this chapter. This article asks the parallel question about operators: is there a family of operators that rotates the way a multiplet of states does? There is, and it is the last structural idea the chapter needs. A spherical tensor operator $T^k_q$ is a set of $2k+1$ operators that shuffle among themselves under rotation with the same $D$ -matrices that shuffle $|k,q\rangle$ states — "operators shaped like spherical harmonics." Why manufacture such objects? Because every observable you care about — dipole moments, quadrupole moments, spin operators, interaction Hamiltonians — must be some assembly of rotational shapes, and matrix elements between $|j,m\rangle$ states depend overwhelmingly on those shapes. Sort an operator into its irreducible ranks and the angular bookkeeping of every matrix element becomes pure Clebsch-Gordan data. That claim — the Wigner-Eckart theorem — is stated here in one line as the payoff and proven in the next article, which owns it. The working content here is the sorting machinery itself. Two equivalent definitions (rotation law and commutator law), proven equivalent by an infinitesimal argument. The rank-1 dictionary converting any Cartesian vector operator — the vector-operator card of Angular Momentum Commutation Relations — into spherical components. The cautionary example: a Cartesian dyadic $A_iB_j$ is reducible, splitting as $9=1\oplus3\oplus5$ into trace, antisymmetric vector, and symmetric-traceless quadrupole parts. And the composition rule: tensors multiply via CG coefficients exactly as states couple. Every identity is machine-verified, including the conventions where sign errors breed.

Key takeaways

A spherical tensor operator of rank $k$ is $2k+1$ operators $T^k_q$ transforming under rotations like the states $|k,q\rangle$ : $\hat{U}^\dagger T^k_q\hat{U}=\sum_{q'}D^{k*}_{qq'}T^k_{q'}$ .
Equivalent commutator definition: $[\hat{J}_z,T^k_q]=q\hbar\,T^k_q$ and $[\hat{J}_\pm,T^k_q]=\hbar\sqrt{k(k+1)-q(q\pm1)}\,T^k_{q\pm1}$ — a $|k,q\rangle$ multiplet in operator clothing.
Rank-1 dictionary for any vector operator: $V^1_0=V_z$ , $V^1_{\pm1}=\mp(V_x\pm iV_y)/\sqrt2$ (Condon-Shortley signs).
Cartesian tensors are reducible: $A_iB_j$ splits as $9=1+3+5$ — scalar $\mathbf{A}\!\cdot\!\mathbf{B}$ , vector $\mathbf{A}\times\mathbf{B}$ , symmetric-traceless quadrupole.
Tensors compose like states: $T^k_q=\sum\langle k_1q_1;k_2q_2|kq\rangle X^{k_1}_{q_1}Z^{k_2}_{q_2}$ — CG coefficients run operator multiplication too.
Payoff (stated, proven next article): matrix elements $\langle\alpha'j'm'|T^k_q|\alpha jm\rangle$ factor into one reduced number times a CG coefficient.
Key idea: operators carry angular momentum — sort any observable into its irreducible ranks and rotations handle all its matrix-element geometry for you.

The core insight

A rank- $k$ spherical tensor operator is a $(2k{+}1)$ -component family $T^k_q$ obeying the same rotation (equivalently, commutation) law as the kets $|k,q\rangle$ ; every observable decomposes into such irreducible ranks, and that decomposition is what makes angular-momentum matrix elements computable.

Where this lives today

Spherical-tensor bookkeeping is the daily language of magnetic resonance: NMR interaction Hamiltonians (chemical shift, dipolar, quadrupolar) are catalogued by rank, and pulse sequences are designed as filters that pass or block specific $T^k_q$ coherences — "multiple-quantum filtering" is rank-and- $q$ selection performed on hardware. In quantum information with high-spin systems, qudit tomography reconstructs the density matrix rank by rank (the $T^k_q$ form an orthogonal operator basis). And nuclear quadrupole resonance, electric-field-gradient calculations in solids, and multipole analyses of gamma decay all run on the $9=1+3+5$ discipline — project first, then compute.

The physical picture

An operator, like a state, has a shape. Apply the dipole operator $\hat{z}$ to an $s$ -state and you get something $p$ -shaped: the operator injected one unit of angular character. Apply the quadrupole $\hat{z}^2-\tfrac13\hat{r}^2$ and you inject two units. "Rank" measures exactly this: a rank- $k$ tensor operator carries $k$ units of angular momentum that it can deposit into (or absorb from) whatever state it touches — the picture that makes selection rules obvious before any theorem is stated.

Figure

Operators carry angular momentum: rank k acts like gluing a k-shaped harmonic onto the state

The wrong basis for this bookkeeping is the Cartesian one. A Cartesian tensor like $A_iB_j$ looks tidy — nine components, two indices — but rotations scramble those nine into reducible junk: part of it rotates like a scalar, part like a vector, part like a genuine rank-2 object. The spherical basis is the one rotations respect:

Figure

A Cartesian dyadic is three irreducible objects wearing one coat: 9 = 1 + 3 + 5

Two definitions, one object

The honest definition mimics the state transformation law. Under a rotation $R$ with unitary $\hat{U}(R)$ , using the $D$ -matrices of Wigner D-Matrices (consumed, not rebuilt):

\hat{U}^\dagger(R)\,T^k_q\,\hat{U}(R) \;=\; \sum_{q'=-k}^{k} D^{k\,*}_{q q'}(R)\; T^k_{q'} \qquad (1)

where: the conjugate and the index order are convention-critical — this exact form (Sakurai's) is pinned numerically in the verifier at $\beta=0.7$ , where the transposed-index variant fails. Read (1) as: rotating the apparatus re-expresses each $T^k_q$ as a $D$ -weighted mix of its $2k+1$ siblings, never leaking into other ranks. The equivalent active form $\hat{U}T^k_q\hat{U}^\dagger=\sum_{q'}D^k_{q'q}T^k_{q'}$ is verified alongside. For computation, the infinitesimal version is worth more. Feeding $\hat{U}=1-i\varepsilon\hat{J}_{\hat{n}}/\hbar$ into (1) and matching first order turns the rotation law into three commutators:

[\hat{J}_z,\,T^k_q]=q\hbar\,T^k_q, \qquad [\hat{J}_\pm,\,T^k_q]=\hbar\sqrt{k(k{+}1)-q(q{\pm}1)}\;T^k_{q\pm1} \qquad (2)

where: compare with $\hat{J}_z|k,q\rangle=q\hbar|k,q\rangle$ and $\hat{J}_\pm|k,q\rangle=\hbar\sqrt{k(k{+}1)-q(q{\pm}1)}|k,q{\pm}1\rangle$ — the adjoint action of the algebra on operators reproduces, coefficient for coefficient, its action on states. That is the precise sense in which $T^k_q$ "is" a $|k,q\rangle$ living in operator space. Equation (2) is also the practical membership test: hand me a candidate family, I compute three commutators and either certify its rank or reject it.

Derivation AEquivalence of the rotation law and the commutator law

We prove definitions (1) and (2) are the same condition. The strategy: expand the finite rotation law to first order in the rotation angle to get the commutators (easy direction), then argue the converse by exponentiation — finite rotations are limits of products of infinitesimal ones, so first-order agreement propagates to the whole group. The series grammar for $e^{-i\varepsilon\hat{J}/\hbar}$ is the translation-operator machinery of The Canonical Commutation Relation and Unitary Operators, consumed here.

Step 1. Infinitesimal rotation about $\hat{n}$ : $\hat{U}=1-\tfrac{i\varepsilon}{\hbar}\hat{J}_{\hat{n}}+O(\varepsilon^2)$ . Left side of (1): $\hat{U}^\dagger T^k_q\hat{U}=T^k_q+\tfrac{i\varepsilon}{\hbar}[\hat{J}_{\hat{n}},T^k_q]+O(\varepsilon^2)$ .

Step 2. Right side of (1): the $D$ -matrix for the same infinitesimal rotation is $D^{k}_{qq'}=\delta_{qq'}-\tfrac{i\varepsilon}{\hbar}\langle k q|\hat{J}_{\hat{n}}|k q'\rangle+O(\varepsilon^2)$ ; conjugating and summing gives $T^k_q+\tfrac{i\varepsilon}{\hbar}\sum_{q'}\langle k q'|\hat{J}_{\hat{n}}|k q\rangle^{\phantom{*}}\!T^k_{q'}$ after using hermiticity of $\hat{J}_{\hat{n}}$ in the $|k,q\rangle$ multiplet.

Step 3. Match orders of $\varepsilon$ : $[\hat{J}_{\hat{n}},T^k_q]=\sum_{q'}T^k_{q'}\langle k q'|\hat{J}_{\hat{n}}|k q\rangle$ . Specialize $\hat{n}$ : for $\hat{J}_z$ the matrix element is $q\hbar\,\delta_{q'q}$ — the first commutator of (2); for $\hat{J}_\pm$ it is the ladder coefficient at $(q',q{\pm}1)$ — the other two. The commutator law is the rotation law read at first order, with the multiplet matrix elements of Ladder Operators and the Angular Momentum Spectrum appearing verbatim. Why the adjoint action inherits the state action: commutation with $\hat{J}$ is how generators act on operators (the Heisenberg-picture grammar), and Step 3 says that action, restricted to the $2k{+}1$ family, has exactly the $|k,q\rangle$ matrix elements. Rank- $k$ tensor operators span an invariant subspace of operator space carrying the same representation as a spin- $k$ multiplet.

Step 4. Converse. Given (2), define the candidate transformation error $E(\varepsilon)=\hat{U}^\dagger T^k_q\hat{U}-\sum D^{k*}_{qq'}T^k_{q'}$ along a one-parameter rotation family. Step 1-3 shows $E'(0)=0$ given (2); differentiating the group composition law shows $E$ obeys a first-order linear ODE with $E(0)=0$ , hence $E\equiv0$ — first-order agreement exponentiates to all angles. Verified: all nine commutators of (2) exact for the rank-1 family built from $\hat{J}$ on $j=1$ ; first-order rotation law checked numerically about two axes; finite law (1) checked at $\beta=0.7$ including the index-order and conjugation conventions — the transposed variant demonstrably fails (article verifier).

The rank-1 dictionary and the reducibility trap

The vector-operator card of Angular Momentum Commutation Relations defined $\hat{\mathbf{V}}$ by $[\hat{J}_i,\hat{V}_j]=i\hbar\varepsilon_{ijk}\hat{V}_k$ . That Cartesian statement is precisely rank-1 in disguise: define the spherical components

V^1_{+1}=-\frac{\hat{V}_x+i\hat{V}_y}{\sqrt2},\qquad V^1_0=\hat{V}_z,\qquad V^1_{-1}=+\frac{\hat{V}_x-i\hat{V}_y}{\sqrt2} \qquad (3)

where: the minus sign on $V^1_{+1}$ is the Condon-Shortley convention reaching into operator space — the same sign that makes $Y_{11}\propto-(x+iy)/r$ . Drop it and half your reduced matrix elements come out with wrong signs when compared against tables. Plugging (3) into the vector commutation relations reproduces (2) with $k=1$ exactly (verified on $\hat{\mathbf{V}}=\hat{\mathbf{J}}$ itself, the always-available example). Position $\hat{\mathbf{r}}$ , momentum $\hat{\mathbf{p}}$ , spin $\hat{\mathbf{S}}$ : every vector operator you own is now a rank-1 spherical tensor via this dictionary. Scalars are rank 0 ( $[\hat{J}_i,T^0_0]=0$ — rotationally inert, one component). The trap begins at two indices:

Go deeper · Derivation B · 1 minDecomposing a Cartesian dyadic into irreducible ranks

Derivation BDecomposing a Cartesian dyadic into irreducible ranks

We prove that the nine-component dyadic $T_{ij}=A_iB_j$ built from two vector operators is reducible, splitting into rank 0 ⊕ 1 ⊕ 2 with dimensions $1+3+5=9$ . The strategy is to exhibit the three invariant blocks explicitly, verify each is closed under rotations, and match the count against the state-coupling series $1\otimes1=0\oplus1\oplus2$ — the same CG arithmetic as coupling two spin-1 states, because that is literally what is happening in operator space.

Step 1. Split algebraically: $A_iB_j=\underbrace{\tfrac{\mathbf{A}\cdot\mathbf{B}}{3}\delta_{ij}}_{\text{trace}}+\underbrace{\tfrac{A_iB_j-A_jB_i}{2}}_{\text{antisymmetric}}+\underbrace{\Big(\tfrac{A_iB_j+A_jB_i}{2}-\tfrac{\mathbf{A}\cdot\mathbf{B}}{3}\delta_{ij}\Big)}_{\text{symmetric traceless}}$ — an identity, verified by reassembly.

Step 2. Identify the ranks. The trace piece is one number times the invariant $\delta_{ij}$ : rank 0. The antisymmetric piece has three independent components and is the cross product in disguise, $\tfrac12(A_iB_j-A_jB_i)=\tfrac12\varepsilon_{ijk}(\mathbf{A}\times\mathbf{B})_k$ : rank 1. The symmetric-traceless remainder has $6-1=5$ components: rank 2, the quadrupole shape.

Step 3. Closure under rotations. Under $T_{ij}\to R_{ia}R_{jb}T_{ab}$ : the trace is invariant ( $R^{\mathsf T}R=1$ ); antisymmetry and symmetry are preserved by the double rotation; the trace of the symmetric part is invariant, so tracelessness survives. Each block maps onto itself — no rank mixing, confirmed numerically with a random rotation matrix. Why "two indices" is not "rank 2": rank is a representation label, not an index count. The dyadic carries the product representation $1\otimes1$ , and products are generically reducible. Physics that depends on rank — selection rules, which multipole a term drives — goes wrong if you feed it an unreduced Cartesian object: the classic error is treating all nine $A_iB_j$ as quadrupole and finding phantom $\Delta j=1$ "quadrupole" transitions that are really the hidden $\mathbf{A}\times\mathbf{B}$ vector at work.

Step 4. Spherical assembly. The same decomposition, CG-organized: $T^k_q=\sum_{q_1q_2}\langle1 q_1;1 q_2|k q\rangle A^1_{q_1}B^1_{q_2}$ for $k=0,1,2$ — tensors compose exactly like states. The verifier builds these and finds $T^0_0=-\tfrac{1}{\sqrt3}\,\mathbf{A}\!\cdot\!\mathbf{B}$ , $T^1_0=\tfrac{i}{\sqrt2}(\mathbf{A}\times\mathbf{B})_z$ , and for $\mathbf{A}=\mathbf{B}=\hat{\mathbf{r}}$ : $T^2_0=\tfrac{2z^2-x^2-y^2}{\sqrt6}$ , $T^2_{\pm2}=\tfrac{(x\pm iy)^2}{2}$ — the quadrupole components, proportional to $r^2Y_{2q}$ as the shape picture promised. Verified: reassembly identity, block counts, rotation closure of all three blocks under a random rotation (to $10^{-13}$ ), and all four CG-assembled component formulas symbolically exact (article verifier).

Note the general composition rule extracted in Step 4 — it is how you manufacture high-rank tensors to order:

T^k_q \;=\; \sum_{q_1,q_2} \langle k_1 q_1; k_2 q_2|k\,q\rangle\, X^{k_1}_{q_1} Z^{k_2}_{q_2} \qquad (4)

where: the CG coefficients of Clebsch-Gordan Coefficients (consumed, in-lane) do double duty: the same table that couples two states couples two operators. With $\mathbf{A}=\mathbf{B}=\hat{\mathbf{r}}$ the antisymmetric rank-1 part vanishes identically ( $\hat{\mathbf{r}}\times\hat{\mathbf{r}}=0$ ), which is why the second-rank multipole family $r^2Y_{2q}$ plus the scalar $r^2$ exhausts the dyadic $\hat{r}_i\hat{r}_j$ — the quadrupole operator of every atomic and nuclear application. And the payoff, stated once and owned next door: because $T^k_q$ carries the $(k,q)$ labels of a state, its matrix elements between $|j,m\rangle$ multiplets inherit CG geometry wholesale,

\langle\alpha' j' m'|T^k_q|\alpha j m\rangle \;=\; \langle j\,m;k\,q|j'\,m'\rangle\,\frac{\langle\alpha'j'\|T^k\|\alpha j\rangle}{\sqrt{2j'+1}} \qquad (5)

where: one reduced number carries all the dynamics; the CG factor carries every $m$ . The proof — the recursion argument this article's base sketched — is relocated to and owned by The Wigner-Eckart Theorem, the next serial, together with the selection-rule corollaries. Here (5) is the destination sign that explains why the sorting machinery was worth building.

Limiting cases

case	result	meaning
$k=0$	$[\hat{J}_i,T^0_0]=0$	scalars are rotation-blind: $\Delta j=\Delta m=0$ matrix structure
$k=1$ , dictionary (3)	vector operators = rank 1	the Cartesian card of the algebra article, re-clothed
$\mathbf{A}=\mathbf{B}=\hat{\mathbf{r}}$ dyadic	rank-1 part vanishes	$\hat{\mathbf{r}}\times\hat{\mathbf{r}}=0$ : pure scalar + quadrupole — the multipole family
$T^k_q$ with $k>2j$ acting within a spin- $j$ space	all matrix elements vanish	the triangle rule closes the window: no rank-3 tensor lives on spin-1
commuting $[\hat{J}^2,T^k_q]$	$\neq0$ in general	tensors shift $j$ : only rank 0 preserves the multiplet

Worked examples

Example 1 — certifying a tensor family. Claim: the spherical components of $\hat{\mathbf{J}}$ itself, $J^1_{\pm1}=\mp(\hat{J}_x\pm i\hat{J}_y)/\sqrt2$ , $J^1_0=\hat{J}_z$ , form a rank-1 tensor. Certify it on the $j=1$ multiplet.

Step 1. Membership test (2), first commutator: $[\hat{J}_z,J^1_q]=q\hbar J^1_q$ — three exact matrix identities (computed: all pass to $10^{-13}$ ).

Step 2. Ladder commutators: $[\hat{J}_+,J^1_{-1}]=\hbar\sqrt2\,J^1_0$ , $[\hat{J}_+,J^1_0]=\hbar\sqrt2\,J^1_{+1}$ , $[\hat{J}_+,J^1_{+1}]=0$ , and mirrors for $\hat{J}_-$ — six more exact identities (computed). The coefficients $\sqrt{2}$ are $\sqrt{k(k{+}1)-q(q{\pm}1)}$ at $k=1$ .

Step 3. Finite-rotation audit: conjugating by $e^{-i(0.7)\hat{J}_y/\hbar}$ reproduces the $d^1(0.7)$ -weighted mix per (1) exactly. Answer: certified rank-1 — and the same nine-commutator test is your general-purpose rank detector for any candidate family.

Example 2 — building the quadrupole operator and reading its reach. Construct $T^2_q$ from the position operator and determine which $\ell\to\ell'$ blocks it can touch in a central-potential atom.

Step 1. Assemble via (4) with $\mathbf{A}=\mathbf{B}=\hat{\mathbf{r}}$ : the verifier's symbolic run gives $T^2_0=(2\hat{z}^2-\hat{x}^2-\hat{y}^2)/\sqrt6=\tfrac{1}{\sqrt6}\,(3\hat{z}^2-\hat{r}^2)$ — the familiar quadrupole shape $\propto 3\hat{z}^2-\hat{r}^2$ — together with $T^2_{\pm2}=(\hat{x}\pm i\hat{y})^2/2$ and $T^2_{\pm1}=\mp(\hat{x}\pm i\hat{y})\hat{z}$ . These are $r^2Y_{2q}$ shapes up to one overall constant.

Step 2. Reach: rank 2 deposits up to two units — matrix elements $\langle\ell'm'|T^2_q|\ell m\rangle$ vanish unless $|\ell-2|\le\ell'\le\ell+2$ and $m'=m+q$ . Parity tightens it further: $T^2$ is parity-even, and $(-1)^{\ell'}=(-1)^{\ell}$ kills $\ell'=\ell\pm1$ — leaving $\Delta\ell\in\{0,\pm2\}$ ( $0\to0$ excluded by the triangle rule).

Step 3. Read a consequence: an electric-quadrupole ( $E2$ ) transition can bridge $s\to d$ ( $\ell:0\to2$ ) but never $s\to p$ — which is why quadrupole lines appear exactly where dipole selection rules said "forbidden." Intensity ratios within the allowed block are CG-squared ratios by (5) — computed in the next article's ratio method. Answer: $T^2_q\propto r^2Y_{2q}$ ; reach $\Delta\ell=0,\pm2$ (no $0\to0$ ), $\Delta m=q$ .

Origins · 1 minUgo Fano and how this was discovered

Ugo Fano

1912–2001 · Rome and Chicago

Fano learned physics in Fermi's Rome group — he was among the last of the ragazzi di via Panisperna — and emigrated in 1939 when the racial laws made Italy impossible. His scientific signature became the systematic algebra of atomic structure: with Giulio Racah he wrote Irreducible Tensorial Sets (1959), the compact book that made spherical tensor operators the standard working language of atomic and collision physics — the state-operator parallel of this article is its organizing principle. At Chicago he built a school (his students ran atomic theory in America for a generation) and left his name on resonance profiles and autoionization lineshapes — a separate story, told by the glossary's Fano-profile entries, not this band. Racah's own wartime half of the partnership belongs to this chapter's closing article.

History

1927

— Wigner's group-theoretic papers organize atomic matrix elements by rotational behavior

1930

— Eckart's vector version: the first "operator multiplet" factorization

1942

— Racah's Theory of Complex Spectra II introduces systematic tensor-operator algebra

1959

— Fano and Racah's Irreducible Tensorial Sets becomes the practitioner's standard

1960

s — Nuclear and atomic many-body theory adopt tensorial methods wholesale (quadrupole moments, EFG tensors, multipole expansions)

Going further

Sakurai & Napolitano §3.11 — definitions, the equivalence argument, and the dyadic example at this article's register.
Fano & Racah, Irreducible Tensorial Sets (1959) — the compact classic; Edmonds Ch. 5 for the formulary.
Messiah Vol. II Ch. XIII; Shankar Ch. 15 — alternative full treatments.

Provenance note: A verifier should: (1) confirm the commutator definitions and their equivalence to the $\mathcal{D}$ -matrix transformation law against Sakurai §3.11; (2) check the rank-1 spherical components $V^1_0=V_z$ , $V^1_{\pm1}=\mp(V_x\pm iV_y)/\sqrt2$ in the Condon–Shortley convention; (3) verify the Wigner–Eckart factorization and the $\sqrt{2j'+1}$ normalization (Sakurai convention) and the resulting $\Delta m=q$ and triangle selection rules; (4) check the Landé g-factor formula $g_J=1+[j(j+1)+s(s+1)-l(l+1)]/[2j(j+1)]$ against any standard atomic-physics text; (5) confirm the dyadic decomposition dimension count $1+3+5=9$ matching $1\otimes1=0\oplus1\oplus2$ . Cross-references: Sakurai & Napolitano Ch. 3; Shankar Ch. 15; Edmonds Ch. 5; Messiah Vol. II Ch. XIII. — Verifier items (1), (2), (5) are executed in this article's verify script. Item (3), the Wigner-Eckart proof, is owned per the chapter map by The Wigner-Eckart Theorem (the base's proof derivation is relocated there — next serial); item (4), the Landé derivation, by Projection Theorem for Vector Operators.

Is a tensor operator a tensor product of operators?

No — unlucky vocabulary collision. "Tensor product" builds composite spaces ( $\mathcal{H}_1\otimes\mathcal{H}_2$ ); a "tensor operator" is a family of operators on one space classified by how rotations shuffle it. The only connection is that you can build tensor operators by CG-coupling two families, equation (4), just as you couple two subsystems' states.

Why do the commutators use the adjoint action rather than ordinary products?

Because conjugation is how transformations act on operators: $T\to\hat{U}^\dagger T\hat{U}$ , whose infinitesimal version is the commutator with the generator. An operator's products with $\hat{J}$ have no transformation meaning by themselves; its commutators are precisely "how the operator changes when the apparatus rotates." The same grammar runs the Heisenberg equation of motion — there with $\hat{H}$ as generator of time, here with $\hat{J}$ as generator of angle.

Are the T^k_q components hermitian observables?

Generally not individually: hermiticity ties $q$ to $-q$ , the standard condition being $(T^k_q)^\dagger=(-1)^qT^k_{-q}$ . Observables are built as real combinations — e.g. $T^2_{+2}+T^2_{-2}\propto\hat{x}^2-\hat{y}^2$ . This mirrors the states: $Y_{\ell m}$ are complex, while real combinations give the $p_x,p_y$ -style orbitals.

Does every operator decompose into spherical tensors?

Yes — the $T^k_q$ families span operator space, so any operator is a sum of pieces of definite rank (possibly infinitely many for unbounded operators, organized by the multipole expansion). The decomposition is the operator analogue of expanding a function on the sphere in $Y_{\ell m}$ : existence is guaranteed; finding the pieces is projection, exactly as in the dyadic derivation.

Why is the minus sign in V¹₊₁ worth caring about?

Because it synchronizes three conventions that must agree for tables to interoperate: the spherical harmonics' CS phases, the CG tables' CS phases, and your operator components. With the sign, $r^1_q=rY_{1q}\sqrt{4\pi/3}$ exactly and the Wigner-Eckart machinery produces table-compatible reduced matrix elements. Without it, every odd-rank reduced element silently flips sign — undetectable until you compare intensities with experiment or someone else's code.

Can rank change under hermitian conjugation or products?

Conjugation preserves rank (with the $q\to-q$ phase above). Products multiply representations: rank $k_1$ times rank $k_2$ contains every rank from $|k_1-k_2|$ to $k_1+k_2$ — the operator triangle rule of equation (4). This is why the square of a dipole operator contains scalar and quadrupole pieces (and why "dipole squared" effects shift energies at rank 0 and rank 2 simultaneously — tensor light shifts in atomic clocks).

What exactly will the Wigner-Eckart theorem add to this article?

The proof and the practice. This article established that $T^k_q$ carries state-like $(k,q)$ labels; the theorem converts that structure into the factorization (5) — geometry (CG factor) times dynamics (one reduced number per $(\alpha'j',\alpha j)$ block) — plus the full selection-rule catalogue and the ratio method for computing everything from one measured element. That machinery, and its recursion-based proof, are the next article's property.