Vector Spherical Harmonics

Twice now this chapter has promised a piece of machinery and deferred it by name. The E9 article Multipole Expansion of Radiation built the ladder of radiation channels and stated that each $(\ell,m)$ carries its angular structure in a transverse vector harmonic $\mathbf X_{\ell m}$ ; the chapter opener Multipole Expansion of Radiation Fields defined the electric and magnetic channels exactly and used the orthogonality of those same objects to collapse the total power into an incoherent sum. Both articles borrowed $\mathbf X_{\ell m}$ on credit. This article pays the debt: it is the algebra of the vector spherical harmonics, done properly. The construction is one line — apply the orbital angular-momentum operator $\mathbf L=-i\,\mathbf r\times\nabla$ to an ordinary spherical harmonic and normalize — but that line is load-bearing. Because $\mathbf L$ is built from a cross product with $\mathbf r$ , the result is automatically tangent to every sphere: exactly the geometry a far-zone radiation field must have. Because $\mathbf L$ generates rotations, $\mathbf X_{\ell m}$ inherits the full angular-momentum quantum numbers of $Y_{\ell m}$ : a single multipole photon really does carry total angular momentum $\ell$ and $z$ -projection $m$ , and the selection rules of atomic and nuclear spectroscopy are this article's orthogonality relations wearing laboratory clothes. The working payoff is concrete. We prove the two orthogonality integrals that make multipole powers add incoherently, establish the sum rule that forces an incoherently populated $\ell$ -level to radiate isotropically, and compute the $\ell=1$ harmonics explicitly — watching the abstract machinery hand back, coefficient for coefficient, the $\sin^2\theta$ donut of Electric Dipole Radiation: Exact Fields and Zone Structure and the four-lobe cloverleaf of the quadrupole article before this one.

Key takeaways

The transverse vector spherical harmonic is $\mathbf X_{\ell m}=\frac{1}{\sqrt{\ell(\ell+1)}}\mathbf L\,Y_{\ell m}$ with $\mathbf L=-i\,\mathbf r\times\nabla$ ; it satisfies $\hat{\mathbf r}\cdot\mathbf X_{\ell m}=0$ identically — tangent to every sphere, born for radiation.
$\{\hat{\mathbf r}Y_{\ell m},\ \mathbf X_{\ell m},\ \hat{\mathbf r}\times\mathbf X_{\ell m}\}$ is a complete orthogonal triad for vector fields on the sphere; radiation uses only the last two.
Two orthogonality relations do all the work: $\int\mathbf X^*_{\ell'm'}\!\cdot\mathbf X_{\ell m}\,d\Omega=\delta_{\ell\ell'}\delta_{mm'}$ and $\int\mathbf X^*_{\ell'm'}\!\cdot(\hat{\mathbf r}\times\mathbf X_{\ell m})\,d\Omega=0$ — same-type modes are orthonormal, and the two polarization families never interfere in totals.
Channel dictionary (the naming trap untangled): an electric multipole has $\mathbf H\propto\mathbf X_{\ell m}$ (TM: radial $\mathbf E$ survives), a magnetic multipole has $\mathbf E\propto\mathbf X_{\ell m}$ (TE: radial $\mathbf H$ survives).
Sum rule: $\sum_m|\mathbf X_{\ell m}|^2=\frac{2\ell+1}{4\pi}$ , a constant — an incoherent, equally populated $\ell$ -level radiates isotropically.
The explicit $\ell=1$ gallery, $|\mathbf X_{10}|^2=\frac{3}{8\pi}\sin^2\theta$ and $|\mathbf X_{1\pm1}|^2=\frac{3}{16\pi}(1+\cos^2\theta)$ , hands back the dipole donut exactly.
Key idea: one operator, $\mathbf L$ , converts the scalar harmonics into the complete transverse basis that diagonalizes radiation into independent TE and TM multipoles with definite angular momentum.

The core insight

Apply the rotation generator $\mathbf L=-i\,\mathbf r\times\nabla$ to $Y_{\ell m}$ and normalize: the resulting $\mathbf X_{\ell m}$ are transverse, orthonormal, angular-momentum-definite vector fields on the sphere, and every radiation field is a sum of $f_\ell(kr)\mathbf X_{\ell m}$ (one polarization) and its curl partner (the other).

Where this lives today

The $\mathbf X_{\ell m}$ alphabet is production machinery across four fields. Nanophotonics designs dielectric metasurfaces by multipole decomposition: scattering simulations are projected onto VSH channels to engineer Kerker-type interference (electric against magnetic dipole terms canceling backscatter) — the design loop of "Mie-tronics" is coefficient bookkeeping in this basis, with the full scattering story deferred by name to the chapter's Mie article. Structured-light laboratories generate beams of definite optical angular momentum whose classical fields are superpositions of small- $\ell$ multipole modes; the statement that a single photon in an $\mathbf X_{\ell m}$ mode carries total angular momentum quantum numbers $(\ell,m)$ is tested directly in such experiments. In cosmology, the CMB community expands polarization maps in the spin-weighted cousins of these harmonics — the E/B decomposition that separates scalar from tensor perturbations is the same "gradient-family versus curl-family" split as $\hat{\mathbf r}\times\mathbf X$ versus $\mathbf X$ . And helioseismology and geophysics decompose oscillations of stars and Earth into spheroidal and toroidal modes — the $\Psi$ - and $\Phi$ -type members of the triad — reading interior structure from which letters of the vector alphabet ring.

Origins · 1 minHermann Weyl and how this was discovered

Hermann Weyl

1885–1955 · Göttingen, Zürich, Princeton — who made symmetry the organizing principle behind bases like this one

Weyl's Gruppentheorie und Quantenmechanik (1928) recast the angular-momentum machinery — the operator $\mathbf L$ , its eigenvalue ladders, the behavior of multiplets under rotation — as representation theory of the rotation group, the framework in which this article's shortcuts live: $\mathbf X_{\ell m}$ transforms like $Y_{\ell m}$ because $\mathbf L$ commutes with rotations, and the sum rule of Derivation C is a two-line consequence of a multiplet's unitary reshuffling. His theorem that the irreducible representations of the rotation group exhaust the possible angular alphabets is why the triad $\{\hat{\mathbf r}Y,\mathbf X,\hat{\mathbf r}\times\mathbf X\}$ is not one convenient choice among many but the canonical decomposition of a vector field on the sphere. Weyl's broader program — gauge invariance (his word), the classification of fields by symmetry — runs through this entire track; here we meet him as the reason a normalized $\mathbf LY_{\ell m}$ was always going to be the right object.

History

1908

— Mie expands light scattering by a sphere in vector partial waves, the first working use of the transverse harmonic alphabet

1928

— Weyl's group-theory treatise systematizes angular momentum as rotation-group representation theory

1935

— Hansen introduces vector wave functions for radiating systems, the direct ancestor of the modern VSH formalism

1952

— Blatt and Weisskopf's nuclear-physics text canonizes multipole fields and transition rates in the X basis

1954

— Hill's American Journal of Physics paper fixes the pedagogical triad notation still in classroom use

The physical picture

A scalar field on a sphere needs one number per point, and the $Y_{\ell m}$ are its natural alphabet. A vector field needs three numbers per point, so it needs three alphabets. One is easy: point radially, with any scalar pattern — that is $\hat{\mathbf r}\,Y_{\ell m}$ . The other two must lie in the sphere's tangent plane, and here geometry offers exactly two natural constructions from a given $Y_{\ell m}$ : walk along the gradient of $Y_{\ell m}$ (downhill on the pattern), or walk along the level curves (around the pattern). The second one — circulation along contour lines — is $\mathbf X_{\ell m}$ up to normalization; the first is its $90°$ -rotated partner $\hat{\mathbf r}\times\mathbf X_{\ell m}$ . Every tangent vector field on the sphere is a combination of these two families. Why single out the contour-circulation family as fundamental? Because it is what the angular-momentum operator produces. $\mathbf L=-i\,\mathbf r\times\nabla$ is the generator of rotations — the same operator, minus its $\hbar$ , that quantum mechanics uses — and acting on $Y_{\ell m}$ it cannot change $\ell$ or $m$ : $\mathbf X_{\ell m}$ transforms under rotations exactly like $Y_{\ell m}$ itself. A radiation mode built on $\mathbf X_{\ell m}$ therefore carries definite total angular momentum — the classical seed of the statement that a multipole photon has $j=\ell$ , and the reason gamma-ray spectroscopists can read nuclear spins from multipolarities.

Figure

The X 1 0 harmonic on the sphere — circulation along lines of latitude with magnitude proportional to sin theta, vanishing at both poles

The triad picture then organizes everything. At each point of the sphere, split any vector field into (radial) + (tangent along $\mathbf X$ ) + (tangent along $\hat{\mathbf r}\times\mathbf X$ ). Far-zone radiation is transverse, so its expansion needs only the two tangent families — and Maxwell's equations couple them in exactly two ways, giving the two multipole types. The channel dictionary deserves care, because half the literature states it sloppily: for an electric multipole the magnetic field is the pure $\mathbf X_{\ell m}$ pattern (so $\mathbf H$ has no radial component anywhere — "transverse magnetic"), while $\mathbf E$ , obtained by a curl, acquires a small radial part in the near zone; for a magnetic multipole the roles swap. The parity tags follow: electric $2^\ell$ -poles carry parity $(-1)^\ell$ , magnetic $2^\ell$ -poles $(-1)^{\ell+1}$ .

Figure

The machinery diagram — one scalar harmonic generates the radial harmonic, the transverse X harmonic, and its rotated partner; Maxwell pairs the two transverse families into TM electric and TE magnetic channels

The core equations

Conventions: SI, harmonics $Y_{\ell m}$ orthonormal on the sphere with the Condon–Shortley phase (as assembled in the E2 article Spherical Harmonics), $\hat{\mathbf r}=\mathbf r/r$ . The orbital angular-momentum operator, dimensionless, is

\mathbf L=-i\,\mathbf r\times\nabla=-i\left(\hat{\boldsymbol\phi}\,\frac{\partial}{\partial\theta}-\hat{\boldsymbol\theta}\,\frac{1}{\sin\theta}\frac{\partial}{\partial\phi}\right) \qquad (1)

where: $\hat{\boldsymbol\theta},\hat{\boldsymbol\phi}$ — tangent unit vectors. Two sanity anchors pin the signs: the $z$ -component works out to $L_z=-i\,\partial/\partial\phi$ (so $L_zY_{\ell m}=mY_{\ell m}$ , the familiar eigenvalue), and $L^2=\mathbf L\cdot\mathbf L$ acting on angles alone satisfies $L^2Y_{\ell m}=\ell(\ell+1)Y_{\ell m}$ . $\mathbf L$ contains no radial derivative: it acts purely on the angular alphabet. The transverse vector spherical harmonic is the normalized image of $Y_{\ell m}$ under this operator:

\mathbf X_{\ell m}(\theta,\phi)=\frac{1}{\sqrt{\ell(\ell+1)}}\,\mathbf L\,Y_{\ell m}(\theta,\phi),\qquad \ell\ge1,\ -\ell\le m\le\ell \qquad (2)

where: the normalization $1/\sqrt{\ell(\ell+1)}$ — fixed by the $L^2$ eigenvalue (Derivation A) so that each $\mathbf X_{\ell m}$ has unit norm on the sphere. For $\ell=0$ the construction returns zero ( $Y_{00}$ is constant): there is no transverse $\ell=0$ harmonic, which is the basis-level statement of the no-monopole-radiation theorem the E9 article proved from charge conservation. Two identities make $\mathbf X_{\ell m}$ the radiation alphabet:

\hat{\mathbf r}\cdot\mathbf X_{\ell m}=0\quad\text{(everywhere)},\qquad \nabla\cdot\mathbf X_{\ell m}=0 \qquad (3)

where: the first is the tangency that far fields require (proved in one line in Derivation A); the second follows from the identity $\nabla\cdot(\mathbf r\times\nabla\psi)=0$ — the $\mathbf X$ family is solenoidal as a three-dimensional field, which is why it can carry a magnetic field all by itself. The two orthogonality relations (Derivation B) are the workhorses:

\int\mathbf X^*_{\ell'm'}\cdot\mathbf X_{\ell m}\,d\Omega=\delta_{\ell\ell'}\delta_{mm'},\qquad \int\mathbf X^*_{\ell'm'}\cdot(\hat{\mathbf r}\times\mathbf X_{\ell m})\,d\Omega=0\ \ \text{for all }\ell,\ell',m,m' \qquad (4)

where: the first makes same-polarization modes orthonormal; the second — valid for every pair, including $\ell=\ell'$ , $m=m'$ — makes the two polarization families non-interfering in every angular integral. Together they are the entire reason multipole powers add incoherently: the cross terms in $|\sum(\text{modes})|^2$ die under $\int d\Omega$ . With the alphabet in hand, the source-free fields outside a radiator expand as (channel dictionary in working form; this is Jackson's expansion, radial functions $f_\ell,g_\ell$ built from spherical Hankel functions for outgoing waves):

\mathbf H=\sum_{\ell m}\Big[a_E(\ell,m)\,f_\ell(kr)\,\mathbf X_{\ell m}-\frac{i}{k}\,a_M(\ell,m)\,\nabla\times\big(g_\ell(kr)\mathbf X_{\ell m}\big)\Big],\quad \mathbf E=Z_0\sum_{\ell m}\Big[\frac{i}{k}\,a_E(\ell,m)\,\nabla\times\big(f_\ell(kr)\mathbf X_{\ell m}\big)+a_M(\ell,m)\,g_\ell(kr)\,\mathbf X_{\ell m}\Big] \qquad (5)

where: $a_E,a_M$ — the electric and magnetic multipole coefficients. Read the structure, not the details: the electric channel puts $\mathbf X_{\ell m}$ into $\mathbf H$ (TM), the magnetic channel puts it into $\mathbf E$ (TE), and duality exchanges them. How $a_E$ and $a_M$ are computed from $\rho$ and $\mathbf J$ — the coefficient integrals with their $j_\ell(kr')$ kernels, their long-wavelength reduction, and the power theorem $P=\frac{Z_0}{2k^2}\sum(|a_E|^2+|a_M|^2)$ — is owned by the chapter opener Multipole Expansion of Radiation Fields and is not repeated here; this article supplies the orthogonality that theorem spends. One more tool, the sum rule (proved in Derivation C by pure symmetry):

\sum_{m=-\ell}^{\ell}\big|\mathbf X_{\ell m}(\theta,\phi)\big|^2=\frac{2\ell+1}{4\pi}\quad\text{— a constant on the sphere} \qquad (6)

where: the content is physical: populate all $2\ell+1$ orientations of an $\ell$ -pole incoherently and equally, and the emission is exactly isotropic — the classical skeleton of why unpolarized atoms with degenerate $m$ -levels show no emission anisotropy.

Derivation ATransversality and the Normalization

We establish the two foundational properties of $\mathbf X_{\ell m}$ : it is tangent to the sphere, and the factor $1/\sqrt{\ell(\ell+1)}$ makes it unit-normalized. Both proofs are short, and both run on operator structure rather than on coordinates — the strategy is to never write a $\theta$ -derivative explicitly, because the algebra of $\mathbf L$ is doing the work. Watch the hermiticity step: it is the same integration-by-parts that quantum mechanics uses, and it is what converts a vector integrand into the scalar eigenvalue problem already solved by the $Y_{\ell m}$ .

Step 1. Tangency. Dot the defining operator with $\mathbf r$ :

\mathbf r\cdot\mathbf L=-i\,\mathbf r\cdot(\mathbf r\times\nabla)=0,

identically zero as a scalar triple product with a repeated vector. Applied to any function, $\mathbf r\cdot\mathbf L Y_{\ell m}=0$ , hence $\hat{\mathbf r}\cdot\mathbf X_{\ell m}=0$ at every point — not asymptotically, not on average: identically. Why this matters structurally: a far-zone field must be tangent to the wavefront sphere; the $\mathbf X$ family satisfies that constraint kinematically, so expanding radiation in it wastes no components.

Step 2. Solenoidality. Using $\nabla\cdot(\mathbf A\times\mathbf B)=\mathbf B\cdot(\nabla\times\mathbf A)-\mathbf A\cdot(\nabla\times\mathbf B)$ with $\mathbf A=\mathbf r$ , $\mathbf B=\nabla Y$ : $\nabla\cdot(\mathbf r\times\nabla Y)=\nabla Y\cdot(\nabla\times\mathbf r)-\mathbf r\cdot(\nabla\times\nabla Y)=0-0=0$ . So $\nabla\cdot\mathbf X_{\ell m}=0$ as a field in three dimensions.

Step 3. Norm. Compute the sphere integral of $|\mathbf L Y_{\ell m}|^2$ using hermiticity of each Cartesian component of $\mathbf L$ on the sphere (they generate rotations, which preserve $d\Omega$ — the same integration-by-parts argument as in quantum mechanics):

\int(\mathbf LY_{\ell m})^*\cdot(\mathbf LY_{\ell m})\,d\Omega=\int Y^*_{\ell m}\,L^2\,Y_{\ell m}\,d\Omega=\ell(\ell+1)\int|Y_{\ell m}|^2\,d\Omega=\ell(\ell+1).

Why hermiticity is legitimate here: each $L_i$ differentiates only angles, and the sphere has no boundary, so the surface terms of the by-parts step vanish — the one hypothesis worth stating out loud.

Step 4. Normalize: dividing by $\sqrt{\ell(\ell+1)}$ gives $\int|\mathbf X_{\ell m}|^2d\Omega=1$ . The same computation read for $\ell=0$ gives zero norm — $\mathbf LY_{00}=0$ — so the transverse alphabet simply has no $\ell=0$ letter: no monopole radiation, now as a statement about the basis itself. Verified: the tangency, solenoidality, and unit norm were checked numerically (quadrature over the sphere) for $(\ell,m)=(1,0),(1,1),(2,0),(2,1)$ : $\langle\mathbf X,\mathbf X\rangle=1$ to $10^{-6}$ , radial components zero to machine precision.

Go deeper · Derivation B · 1 minBoth Orthogonality Relations

Derivation BBoth Orthogonality Relations

We prove the two relations (4). The first — same-family orthonormality — is a two-line extension of the norm computation. The second — the vanishing of every $\mathbf X\cdot(\hat{\mathbf r}\times\mathbf X)$ integral — is the one that usually gets waved through, so we do it honestly; the route is to recognize $\hat{\mathbf r}\times\mathbf X_{\ell m}$ as the gradient-type tangent harmonic and reduce the mixed integral to the flux of a curl through a closed surface, which vanishes by Stokes's theorem. This second relation is what forbids TE–TM interference in any total.

Step 1. Same-family orthogonality. Repeat Step 3 of Derivation A with different indices:

\int(\mathbf LY_{\ell'm'})^*\cdot(\mathbf LY_{\ell m})\,d\Omega=\int Y^*_{\ell'm'}L^2Y_{\ell m}\,d\Omega=\ell(\ell+1)\,\delta_{\ell\ell'}\delta_{mm'},

by hermiticity and then scalar-harmonic orthonormality (the E2 article's result, recalled). Dividing by the two normalizations gives $\int\mathbf X^*_{\ell'm'}\cdot\mathbf X_{\ell m}\,d\Omega=\delta_{\ell\ell'}\delta_{mm'}$ .

Step 2. Identify the rotated partner. Compute $\hat{\mathbf r}\times\mathbf L$ acting on an angular function: with $\hat{\mathbf r}\times(\mathbf r\times\nabla)=\mathbf r\,(\hat{\mathbf r}\cdot\nabla)-r\nabla$ and $(\hat{\mathbf r}\cdot\nabla)Y=0$ for $r$ -independent $Y$ ,

\hat{\mathbf r}\times\mathbf LY=i\,r\nabla Y=i\left(\hat{\boldsymbol\theta}\,\partial_\theta+\hat{\boldsymbol\phi}\,\tfrac{1}{\sin\theta}\partial_\phi\right)Y.

So $\hat{\mathbf r}\times\mathbf X_{\ell m}$ is (up to $i$ and the normalization) the tangential gradient of $Y_{\ell m}$ : the downhill family, rotated $90°$ from the contour family, exactly as the physical picture promised.

Step 3. Reduce the mixed integral. Using Step 2,

\int\mathbf X^*_{\ell'm'}\cdot(\hat{\mathbf r}\times\mathbf X_{\ell m})\,d\Omega\ \propto\ \int(\mathbf r\times\nabla Y^*_{\ell'm'})\cdot(r\nabla Y_{\ell m})\,d\Omega=\int r^2\,\hat{\mathbf r}\cdot\big(\nabla Y^*_{\ell'm'}\times\nabla Y_{\ell m}\big)\,d\Omega,

after cycling the scalar triple product $(\mathbf r\times\nabla Y^*)\cdot\nabla Y=\mathbf r\cdot(\nabla Y^*\times\nabla Y)$ .

Step 4. Recognize a curl and close the surface. For any two scalars, $\nabla Y^*\times\nabla Y=\nabla\times(Y^*\nabla Y)$ . The integral is therefore the flux of a curl through the closed sphere:

\oint\big[\nabla\times(Y^*_{\ell'm'}\nabla Y_{\ell m})\big]\cdot d\mathbf a=0,

which vanishes identically — a closed surface has no boundary for Stokes's theorem to feed. This holds for every index pair, including equal indices: the two tangent families are pointwise orthogonal in the diagonal case and integrally orthogonal always. Why the physics cares: expand a general radiation field in both families and square for the power; relation (ii) kills every electric-magnetic cross term and relation (i) diagonalizes the rest — the incoherent power sum $P\propto\sum(|a_E|^2+|a_M|^2)$ of the chapter opener is these two integrals and nothing more. *Verified: numerically, $\langle\mathbf X_{10},\mathbf X_{20}\rangle<10^{-10}$ , $\langle\mathbf X_{10},\hat{\mathbf r}\times\mathbf X_{10}\rangle=0$ exactly, $\langle\mathbf X_{21},\hat{\mathbf r}\times\mathbf X_{21}\rangle<10^{-11}$ by quadrature.*

Go deeper · Derivation C · 2 minThe Explicit Ladder at l Equals One and the Sum Rule

Derivation CThe Explicit Ladder at l Equals One and the Sum Rule

We compute the three $\ell=1$ harmonics explicitly and prove the sum rule (6). The goal is the dictionary between the abstract alphabet and the patterns this chapter has already met: $\mathbf X_{10}$ must be the electric dipole's donut, and the $m$ -summed level must be isotropic. The route for the sum rule is pure symmetry — no integrals, no special functions — the kind of argument worth internalizing because it generalizes to every $\ell$ .

Step 1. $\mathbf X_{10}$ . With $Y_{10}=\sqrt{3/4\pi}\cos\theta$ , only the $\hat{\boldsymbol\phi}\,\partial_\theta$ term of (1) acts:

\mathbf LY_{10}=-i\hat{\boldsymbol\phi}\,\partial_\theta Y_{10}=i\sqrt{\tfrac{3}{4\pi}}\sin\theta\,\hat{\boldsymbol\phi}\ \Rightarrow\ \mathbf X_{10}=\frac{\mathbf LY_{10}}{\sqrt2}=i\sqrt{\tfrac{3}{8\pi}}\,\sin\theta\,\hat{\boldsymbol\phi},

so $|\mathbf X_{10}|^2=\frac{3}{8\pi}\sin^2\theta$ : azimuthal circulation, zeros at the poles, the dipole donut. For an electric dipole along $z$ , the exact article's far field had $\mathbf H$ azimuthal with amplitude $\propto\sin\theta$ — the $(\ell,m)=(1,0)$ electric channel, letter for letter.

Step 2. $\mathbf X_{1,\pm1}$ . With $Y_{1,\pm1}=\mp\sqrt{3/8\pi}\sin\theta\,e^{\pm i\phi}$ , both terms of (1) act, giving

\mathbf X_{1,\pm1}=\mp\sqrt{\tfrac{3}{16\pi}}\,e^{\pm i\phi}\big(\hat{\boldsymbol\theta}\ \pm\ i\cos\theta\,\hat{\boldsymbol\phi}\big)\cdot(\pm1)\ \Rightarrow\ |\mathbf X_{1,\pm1}|^2=\frac{3}{16\pi}\,(1+\cos^2\theta).

These are the rotating-dipole patterns: maximum along the poles ( $1+\cos^2=2$ ), half-strength in the equator — the angular fingerprint of circular polarization about $z$ , and (recalling the previous article by name) exactly the pattern a rotating charge pair's quadrupole pair-partner, the rotating dipole, would emit at $\Omega$ . Why $m=\pm1$ peaks where $m=0$ vanishes: $m$ counts angular momentum about $z$ carried away along $\hat{\mathbf z}$ ; a circularly rotating source communicates its handedness most efficiently along the rotation axis, while a linear oscillation cannot radiate along its own line.

Step 3. The $\ell=1$ sum. Add the three:

\sum_{m=-1}^{1}|\mathbf X_{1m}|^2=\frac{3}{8\pi}\sin^2\theta+2\cdot\frac{3}{16\pi}(1+\cos^2\theta)=\frac{3}{8\pi}\big(\sin^2\theta+1+\cos^2\theta\big)=\frac{3}{4\pi}.

Constant — the incoherent, equally weighted $\ell=1$ level radiates isotropically.

Step 4. The sum rule for every $\ell$ , by symmetry. The function $F(\theta,\phi)=\sum_m|\mathbf X_{\ell m}|^2$ is built from the complete $m$ -multiplet, so a rotation of the sphere merely reshuffles the $m$ -basis by a unitary matrix and leaves $F$ invariant: $F$ is a rotation-invariant function on the sphere, hence a constant. Its value follows from the norms: $\oint F\,d\Omega=\sum_m1=2\ell+1$ , so $F=(2\ell+1)/4\pi$ . No integrals were harmed. Verified: all component formulas and both sum rules were checked numerically ( $|\mathbf X_{10}|^2$ , $|\mathbf X_{1\pm1}|^2$ , $|\mathbf X_{20}|^2=\frac{15}{8\pi}\sin^2\theta\cos^2\theta$ evaluated at multiple angles; $\sum_m|\mathbf X_{1m}|^2=3/4\pi$ and $\sum_m|\mathbf X_{2m}|^2=5/4\pi$ constant over the sphere to $10^{-9}$ away from the coordinate poles).

Limiting cases

case	result	meaning
$\ell=0$	$\mathbf LY_{00}=0$ : no transverse harmonic	No monopole radiation — the basis lacks the letter; charge-conservation proof in the E9 ladder article
$(\ell,m)=(1,0)$ electric channel	$\mathbf H\propto\mathbf X_{10}$ , $\|\mathbf X_{10}\|^2\propto\sin^2\theta$	The exact-dipole article's far field recovered from pure algebra
$(\ell,m)=(1,\pm1)$	$\|\mathbf X_{1\pm1}\|^2\propto1+\cos^2\theta$	Rotating-dipole (circular polarization) pattern, peaked on the axis
$\ell=2$ , $m=0$	$\|\mathbf X_{20}\|^2=\frac{15}{8\pi}\sin^2\theta\cos^2\theta$	The axial quadrupole cloverleaf of the previous article, as one basis letter
incoherent full $m$ -level	$\sum_m\|\mathbf X_{\ell m}\|^2=\frac{2\ell+1}{4\pi}$	Isotropic emission from unpolarized degenerate levels
far zone of expansion (5)	$f_\ell\to(-i)^{\ell+1}e^{ikr}/kr$	Every multipole becomes an outgoing spherical wave dressed by its $\mathbf X_{\ell m}$

Common trapTransverse to the Radius, Not to the z Axis — and TE Is Not "the Electric Multipole"

Two naming traps sink more homework than the algebra does. First: the "transverse" in these harmonics means $\hat{\mathbf r}\cdot\mathbf X_{\ell m}=0$ — tangent to the sphere — not transverse to some fixed Cartesian axis as for plane waves; the two notions only merge locally in the far zone where the propagation direction is $\hat{\mathbf r}$ itself. Second: the labels pair up crosswise. TE ("transverse electric," $\mathbf E\propto\mathbf X$ , no radial $\mathbf E$ ) is the magnetic multipole; TM ( $\mathbf H\propto\mathbf X$ , no radial $\mathbf H$ ) is the electric multipole. The multipole name tracks which source moment drives the mode; the TE/TM name tracks which field is purely tangent. Books that blur this emit sign-correct formulas under swapped labels — check the radial components, not the names.

Worked examples

Example 1 — the $\ell=1$ level, numerically (confidence builder). Evaluate the $\ell=1$ patterns at $\theta=0°,45°,90°$ and verify isotropy of the incoherent level.

Step 1. $|\mathbf X_{10}|^2=\frac{3}{8\pi}\sin^2\theta$ : values $0$ , $0.0597$ , $0.1194\ \mathrm{sr^{-1}}$ at $0°,45°,90°$ .

Step 2. $|\mathbf X_{1\pm1}|^2=\frac{3}{16\pi}(1+\cos^2\theta)$ : values $0.1194$ , $0.0895$ , $0.0597\ \mathrm{sr^{-1}}$ each.

Step 3. Sum over the three states: $0+2(0.1194)=0.2387$ ; $0.0597+2(0.0895)=0.2387$ ; $0.1194+2(0.0597)=0.2387$ — constant, equal to $3/4\pi=0.2387\ \mathrm{sr^{-1}}$ . An unpolarized dipole level radiates isotropically, angle by angle.

Step 4. Read the physics: a gas of excited atoms with all three $m$ -sublevels equally populated shows no emission anisotropy; apply a magnetic field to lift the degeneracy and observe along $\hat{\mathbf z}$ , and the $m=0$ (π) line vanishes while the $m=\pm1$ (σ) lines survive — the Zeeman-geometry rules are rows of this table.

Example 2 — reading a nuclear E2 pattern (exam level). A nucleus decays by an electric quadrupole transition with $\Delta m=0$ (the $(\ell,m)=(2,0)$ channel). Detectors sit at $\theta=30°,45°,60°,90°$ . Predict the relative count rates and the polarization character.

Step 1. The angular distribution of a pure $(2,0)$ electric multipole is $|\mathbf X_{20}|^2=\frac{15}{8\pi}\sin^2\theta\cos^2\theta$ (transversality makes $|\hat{\mathbf r}\times\mathbf X|^2=|\mathbf X|^2$ , so TM vs TE does not change the intensity pattern — only the polarization).

Step 2. Values: $\sin^2\theta\cos^2\theta=0.1875$ at $30°$ , $0.25$ at $45°$ , $0.1875$ at $60°$ , $0$ at $90°$ . Relative rates $0.75:1:0.75:0$ — the cloverleaf: maximum at $45°$ , zeros on both the axis and the equator, the four-lobe signature that distinguishes E2 from any dipole pattern (which never has an equatorial zero).

Step 3. Absolute scale check at the maximum: $|\mathbf X_{20}(45°)|^2=\frac{15}{8\pi}(0.25)=0.1492\ \mathrm{sr^{-1}}$ , matching the numeric evaluation in Derivation C's verification.

Step 4. Polarization: electric multipole means $\mathbf H\propto\mathbf X_{20}$ — at the $45°$ maximum, $\mathbf X_{20}\propto\sin\theta\cos\theta\,\hat{\boldsymbol\phi}$ is azimuthal, so $\mathbf H$ is azimuthal and $\mathbf E$ lies in the meridional plane: the same polarization geometry as E1, but with the cloverleaf intensity map. Measuring both the zeros and the polarization pins the transition as E2 with $\Delta m=0$ — this is how multipolarities are assigned in gamma spectroscopy.

Going further

References. Jackson, Classical Electrodynamics, 3rd ed., Ch. 9 (esp. §9.7, §9.10) and Ch. 16 · Griffiths, Introduction to Electrodynamics, 4th ed., Ch. 11 · Zangwill, Modern Electrodynamics, Ch. 21 · Morse & Feshbach, Methods of Theoretical Physics, Vol. II, Ch. 13 · Hill, 'The Theory of Vector Spherical Harmonics', Am. J. Phys. 22, 211 (1954) · Blatt & Weisskopf, Theoretical Nuclear Physics, Appendix B

Provenance. Definitions and normalization follow Jackson 3rd ed. §9.7–§9.10 (eqs. for $\mathbf X_{\ell m}$ , multipole fields, power). A verifier should: (1) confirm $\mathbf r\cdot(\mathbf r\times\nabla)=0$ and $L^2 Y_{\ell m}=\ell(\ell+1)Y_{\ell m}$ ; (2) recompute the orthonormality integral; (3) check the $\ell=1$ electric multipole reduces to the $\sin^2\theta$ dipole pattern and Larmor power $P=\mu_0 p_0^2\omega^4/(12\pi c)$ (Griffiths Ch. 11); (4) verify the far-zone Hankel asymptotics $h_\ell^{(1)}(kr)\to(-i)^{\ell+1}e^{ikr}/(kr)$ . Standard values of $Y_{10}=\sqrt{3/4\pi}\cos\theta$ and $Z_0\approx376.73\ \Omega$ are textbook constants. Alternate Hill basis ( $\mathbf Y,\boldsymbol\Psi,\boldsymbol\Phi$ ) cross-checked against Morse & Feshbach Vol. II. Addendum (E11 upgrade, base errors fixed): the base's spherical-coordinate expression for $\mathbf L$ carried an overall sign inconsistent with $L_z=-i\partial_\phi$ and with its own $\mathbf X_{10}$ (corrected in eq. 1); its TE/TM field equations assigned $\mathbf B\propto\mathbf X$ to the magnetic multipole (the assignment is the reverse — see the channel dictionary and warning box) and contained a garbled curl expression, both repaired against Jackson 9.122; and the $m$ -summed level constant is $(2\ell+1)/4\pi$ , verified numerically for $\ell=1,2$ along with all explicit component formulas.

Builds on: Multipole Expansion of Radiation Fields · Electric Dipole Radiation · Magnetic Dipole and Electric Quadrupole Radiation Leads to: Center-Fed Linear Antenna · Antenna Arrays and Directivity · Synchrotron Radiation

How does this article divide labor with the chapter opener on multipole expansion?

The opener Multipole Expansion of Radiation Fields owns the physics pipeline: defining the exact E $\ell$ /M $\ell$ channels from the radial generating scalars, the coefficient integrals $a_E(\ell,m)$ , $a_M(\ell,m)$ with their $j_\ell$ kernels, the long-wavelength reduction, and the power/angular-momentum theorems. This article owns the alphabet those results are written in: the construction, transversality, both orthogonality relations, the sum rule, and the explicit low- $\ell$ gallery. The opener quotes the orthogonality; here it is proved.

Are the X harmonics related to the gradient-and-curl vector harmonics (the Hill Psi and Phi basis)?

Linearly and completely. Hill's triad is $\mathbf Y_{\ell m}=\hat{\mathbf r}Y$ , $\boldsymbol\Psi_{\ell m}=r\nabla Y$ (the downhill family), $\boldsymbol\Phi_{\ell m}=\mathbf r\times\nabla Y$ (the contour family). Our $\mathbf X_{\ell m}$ is $\boldsymbol\Phi_{\ell m}$ normalized (times $-i$ ), and $\hat{\mathbf r}\times\mathbf X_{\ell m}$ is $\boldsymbol\Psi_{\ell m}$ normalized (times $i$ ) — Derivation B's Step 2 is exactly this identification. Radiation theory prefers the $\mathbf X$ convention because the angular-momentum quantum numbers ride explicitly; seismology prefers Hill's names (toroidal = $\Phi$ -type, spheroidal = $\mathbf Y$ + $\Psi$ -type).

Why does the photon of a multipole mode have total angular momentum l rather than l plus spin?

Because $\mathbf X_{\ell m}$ is not an orbital eigenfunction times a fixed polarization — the vector direction rotates with position, entangling orbital and spin parts. The object transforms under rotations as a single irreducible $(\ell,m)$ multiplet (it is $\mathbf L Y_{\ell m}$ , and $\mathbf L$ commutes with total rotations), so the total angular momentum is $\ell$ : the split into "orbital plus spin" is basis-dependent and not sharp for photons, while total $j=\ell$ , $j_z=m$ , and parity are the good quantum numbers. That is precisely the labeling scheme of nuclear gamma transitions.

Do the X family and its rotated partner really span every radiation field, or could something be missed?

Together with the radial family they span all square-integrable vector fields on the sphere — completeness is inherited from the scalar $Y_{\ell m}$ , since the three families are built by applying the three pointwise-independent operations (radial projection, contour flow, gradient flow) to a complete scalar set. For radiation (source-free, transverse at infinity) the radial family's coefficients are fixed by the transverse ones through Maxwell's equations, so the two tangent families with radial Hankel dressings capture everything — that statement, mode function by mode function, is the expansion (5).

Why is the normalization 1 over the square root of l times l plus 1 rather than something with 2l+1?

Because the norm of $\mathbf LY_{\ell m}$ is set by the $L^2$ eigenvalue, not by the multiplicity: $\int|\mathbf LY|^2=\langle Y|L^2|Y\rangle=\ell(\ell+1)$ . The $2\ell+1$ appears elsewhere — in the sum rule over the full multiplet — and confusing the two is a quick route to wrong prefactors. Mnemonic: one mode's normalization knows only its eigenvalue; the level's isotropy constant knows the multiplicity.

What changes for the CMB E and B modes if the machinery is 'the same split'?

The CMB polarization field is a spin-2 quantity (a headless tensor on the sky), so the literal $\mathbf X_{\ell m}$ (spin-1) alphabet is replaced by spin-weighted spherical harmonics — but the structural move is identical: decompose the tangent field into a gradient-type family (E modes, parity $(-1)^\ell$ ) and a curl-type family (B modes, parity $(-1)^{\ell+1}$ ), the exact analogues of $\hat{\mathbf r}\times\mathbf X$ and $\mathbf X$ . The physics payoff mirrors ours too: the two families do not interfere in power spectra, and their different parities let cosmologists separate scalar from tensor (gravitational-wave) sources.

When I look up these formulas in another book the signs of L and X differ — who is right?

Possibly everyone: three sign conventions circulate (the sign of $i$ in $\mathbf L$ , the Condon–Shortley phase inside $Y_{\ell m}$ , and occasional definitions of $\mathbf X$ with an extra $-i$ ). All observables — $|\mathbf X|^2$ patterns, orthogonality, powers — are convention-blind. The self-consistency checks worth running on any source: $L_z=-i\partial_\phi$ must give $L_zY_{\ell m}=+mY_{\ell m}$ , and $\mathbf X_{10}$ must come out azimuthal with $|\mathbf X_{10}|^2=\frac{3}{8\pi}\sin^2\theta$ . A source failing its own checks (as this article's base draft did) has an internal sign slip, not a different convention.