Wigner D-Matrices

Wherever an oriented quantum system is rotated, produced, or decays, a d-function is on duty. Particle physicists fit resonance spins by matching decay angular distributions to sums of $|d^{j}_{\lambda\lambda'}(\theta)|^2$ — the helicity formalism of every amplitude-analysis paper from LHCb or BESIII. Molecular spectroscopists use D-functions as the actual eigenfunctions of the symmetric top ($\psi_{JKM}\propto D^{J*}_{MK}$), so microwave line strengths are Peter–Weyl integrals. And the spin-1 example above is everyday calibration in nitrogen-vacancy magnetometry and cold-atom labs: tilt the axis, read populations — the $\tfrac14,\tfrac12,\tfrac14$ row is the first sanity check that your rotations are what you think they are.

Picture a quantum top of fixed spin length $j$: reorient it any way you like, you can never change how much angular momentum it has — only which way it points. In matrix language the rotation operator is block-diagonal, one block per $j$, because $[\hat U,\hat{\mathbf J}^2]=0$. The D-matrix is one block, written in the $|j,m\rangle$ basis Ladder Operators and the Angular Momentum Spectrum constructed.

The second picture is the factorization. Rotations about the quantization axis are free: they rephase each $|j,m\rangle$ by $e^{-im\times\text{angle}}$ — the operator face of the azimuthal winding $e^{im\phi}$. Only tilting the axis does real work. The Euler decomposition quarantines the work: phase comb, one tilt, phase comb. Whatever question you ask of a rotation, the hard part is a single real function of one angle, $d^{j}_{m'm}(\beta)$.

Half-integer $j$ keeps its double-cover signature: $d^{1/2}(2\pi)=-\mathbb 1$, so these blocks represent $SU(2)$, not $SO(3)$ — the previous article's theorem, now visible entry by entry. Probabilities $|D|^2$ are single-valued either way.

For a rotation $R$, the block of the rotation operator on the $j$-multiplet is

where: $m$ labels the input rung, $m'$ the output; unitarity of $\hat U$ makes each block unitary, and $[\hat U,\hat{\mathbf J}^2]=0$ makes the blocks exact. Parametrize $R$ by Euler angles in the active z-y-z convention:

where: read as space-fixed operations right-to-left — first turn by $\gamma$ about $\hat{\mathbf z}$, then tilt by $\beta$ about the space-fixed $\hat{\mathbf y}$, then turn by $\alpha$ about $\hat{\mathbf z}$. The equivalent body-frame reading runs in the opposite order about carried axes ($\alpha$ about $z$, then $\beta$ about the new $y'$, then $\gamma$ about the newest $z''$) — the two readings give the same operator, a standard identity of rotation kinematics, but mixing them mid-calculation is the classic Euler blunder. Textbooks differ on all of it: z-y-z versus z-x-z, active versus passive, sign of the exponent. Everything in this article is pinned to (2); translate before comparing tables.

The spin-½ block. With $\hat J_y=\tfrac{\hbar}{2}\sigma_y$, the exponential identity owned by Spin-1/2 and the Pauli Matrices — series split on $(\hat{\mathbf n}\cdot\boldsymbol\sigma)^2=\mathbb 1$, consumed here, not re-derived — evaluates the tilt in one line:

where: rows and columns are ordered $m',m=+\tfrac12,-\tfrac12$; the matrix is real and orthogonal, the half-angle announces the double cover ($d^{1/2}(2\pi)=-\mathbb 1$), and $|d^{1/2}_{m'm}|^2$ reproduces every tilted Stern–Gerlach probability — $\cos^2(\beta/2)$ for spin-up to stay spin-up, the number Spinors and Two-Component Wavefunctions derived from eigenspinor overlaps and The Stern-Gerlach Experiment measures.

The spin-1 block, built by the ladder.

Every rung at once: Wigner's closed form. For arbitrary $j$ the tilt matrix is a finite factorial sum,

where: $k$ runs over the integers keeping every factorial argument non-negative, and the sign factor is $(-1)^{k-m+m'}$ — not $(-1)^{k}$. ⚠ The AllQFT export this article upgrades printed $(-1)^{k}$ with these same factorials and exponents; that version is machine-checkably wrong — it reproduces the transpose $d^{j}_{mm'}$, silently flipping every element with odd $m'-m$, and contradicts the export's own $d^{1/2}$ (its off-diagonal $-\sin\tfrac\beta2$ becomes $+\sin\tfrac\beta2$). The corrected sign restores agreement with the exponential definition for every element at $j=\tfrac12,1,\tfrac32$. One more consistency reading: at $m=j$ only $k=0$ survives and (4) collapses to the stretched-column law of the derivation above.

The D-matrix elements are not just tables — they are a complete orthogonal function system on the rotation group, the group-manifold analogue of $e^{im\phi}$ on the circle. For integer $j$, integrating over all rotations with the invariant (Haar) measure $\mathrm dR=\mathrm d\alpha\,\sin\beta\,\mathrm d\beta\,\mathrm d\gamma$, $\alpha,\gamma\in[0,2\pi]$, $\beta\in[0,\pi]$:

where: $8\pi^2$ is the volume of $SO(3)$ in these coordinates, and the $1/(2j+1)$ is Peter–Weyl's general theorem for compact groups: each matrix element of each irreducible representation, weighted by dimension, is one unit vector of an orthogonal basis for functions on the group. Two honest fine-prints. First, as stated (5) is the $SO(3)$ theorem: it covers integer $j$, which is what the spherical-harmonic connection uses. Half-integer representations live on $SU(2)$, whose Haar volume is $16\pi^2$ (extend $\gamma$ to $4\pi$); on that doubled range (5) holds for all $j$ with $16\pi^2$ in place of $8\pi^2$. Second, the doubling is not pedantry: on the un-doubled $2\pi$ ranges, a half-integer and an integer element are not orthogonal — the cross integral $\int D^{1/2*}_{\frac12\frac12}D^{1}_{00}\,\mathrm dR$ is numerically nonzero, and only the $4\pi$ range kills it. (The export stated (5) for general $j_1,j_2$ on the $2\pi$ ranges and labelled $8\pi^2$ the $SU(2)$ volume; both slips are corrected here and pinned in the verify script.) The composition law $D^{j}(R_1)D^{j}(R_2)=D^{j}(R_1R_2)$ — each block inherits the group multiplication — is Rotation Operators and SU(2)'s homomorphism, recalled, and numerically spot-checked rather than re-proved. Products of D-matrices belonging to different systems decompose as sums over total angular momentum — the Clebsch–Gordan series

where: the coefficients are the Clebsch–Gordan coefficients with $M=m_1+m_2$, $M'=m_1'+m_2'$; this is the rotation-matrix face of coupling, and it is owned downstream — the decomposition theorem by Addition of Angular Momenta, the coefficient machinery by Clebsch-Gordan Coefficients. Here it is stated so you recognize the D-matrices as the loom the coupling articles weave on.

Sometimes only the rotation angle matters. Tracing the block gives the character; the trace is basis-free, so trace the diagonal phases of a $z$-rotation by the same angle:

where: the second equality is a finite geometric sum (machine-checked symbolically for $j=\tfrac12,1,\tfrac32$). Characters are the fingerprint tools of representation theory: $\chi^{j}(2\pi)=(2j+1)(-1)^{2j}$ re-announces the double cover, and integrating products of characters counts how many times each $J$ appears in a product — the fast route to the Clebsch–Gordan series' bookkeeping.

Example 1 — the tilted analyzer, three ways (confidence builder). Task: a spin-½ particle exits an SG magnet spin-up along $\hat{\mathbf z}$ and meets an analyzer tilted by $\beta=60^\circ$. Find the probability it emerges spin-up along the new axis.

Step 1. The amplitude is a single little-$d$ element: $d^{1/2}_{\frac12\frac12}(60^\circ)=\cos30^\circ=\tfrac{\sqrt3}{2}$.

Step 2. Square it: $P=\cos^2(\beta/2)=\cos^230^\circ=\boldsymbol{0.75}$.

Step 3. Cross-check the pedigree: Spinors and Two-Component Wavefunctions got $\cos^2(\beta/2)$ from eigenspinor overlaps, The Stern-Gerlach Experiment uses it for cascaded magnets — three formalisms, one number. The complementary channel takes $\sin^2 30^\circ=0.25$, and the two sum to 1 because the $d$-column is a unit vector. $P=0.75$.

Example 2 — tilting a spin-1 atom (exam level). Task: an atom is prepared in $|1,+1\rangle$ (say, optically pumped along $\hat{\mathbf z}$). Its quantization axis is then analyzed along a direction tilted by $\beta=90^\circ$. Find the populations of $m'=+1,0,-1$, and repeat for $\beta=180^\circ$.

Step 1. The amplitudes are the first column of $d^{1}(\beta)$: $\big(\tfrac{1+\cos\beta}{2},\ \tfrac{\sin\beta}{\sqrt2},\ \tfrac{1-\cos\beta}{2}\big)$.

Step 2. At $\beta=90^\circ$: amplitudes $\big(\tfrac12,\tfrac1{\sqrt2},\tfrac12\big)$, populations $\boldsymbol{\tfrac14,\ \tfrac12,\ \tfrac14}$ — half the population lands in $m'=0$, and the original stretched state keeps only a quarter. Check: $\tfrac14+\tfrac12+\tfrac14=1$.

Step 3. At $\beta=180^\circ$: $\cos\beta=-1$, $\sin\beta=0$, so the column is $(0,0,1)$ — the atom arrives entirely in $m'=-1$. An upside-down analyzer sees the stretched state stretched the other way, with certainty; no smearing, because $180^\circ$ maps the $z$-axis onto itself.

Step 4. The classical echo: for large $j$ the analogous corner element $\cos^{2j}(\beta/2)$ at $\beta=90^\circ$ is $2^{-j}$ — a macroscopic spin essentially never survives a right-angle tilt in its stretched state, which is why classical gyroscopes have definite directions. Populations $\tfrac14,\tfrac12,\tfrac14$ at $90^\circ$; total inversion at $180^\circ$.

References. Sakurai & Napolitano, Modern Quantum Mechanics, 3rd ed., Ch. 3 (Theory of Angular Momentum, §3.5–3.9) · Shankar, Principles of Quantum Mechanics, 2nd ed., Ch. 12 (Rotational Invariance and Angular Momentum) · Cohen-Tannoudji, Diu & Laloe, Quantum Mechanics, Vol. 1, Complement B-VI; Vol. 2 · Edmonds, Angular Momentum in Quantum Mechanics, Ch. 4 · Varshalovich, Moskalev & Khersonskii, Quantum Theory of Angular Momentum · Wigner, Group Theory and its Application to the Quantum Mechanics of Atomic Spectra (1959) · Schwinger, "On Angular Momentum," AEC report NYO-3071 (1952)

Provenance. Definitions and factorization follow Sakurai & Napolitano Ch. 3 (z-y-z convention); the spin-1/2 D-matrix matches Sakurai Eq. (3.5.51) and standard Pauli exponentiation. Wigner's explicit little-d formula and the $d^{1/2},d^{1}_{00}=\cos\beta$ values are standard (Edmonds Ch. 4; Varshalovich). Verifier should: (1) expand $e^{-i\sigma_y\beta/2}$ and confirm the $j=1/2$ matrix; (2) check $d^{1}_{00}(\beta)=\cos\beta$ and $d^{j}_{jj}=\cos^{2j}(\beta/2)$ from Wigner's sum; (3) confirm orthogonality normalization $8\pi^2/(2j+1)$ against the Haar measure; (4) verify the spherical-harmonic reduction $D^{\ell}_{m0}=\sqrt{4\pi/(2\ell+1)}\,Y_\ell^{m*}$. Cross-check Shankar Ch. 12 and Cohen-Tannoudji Complement B-VI. Edition note: two corrections to the export were made and machine-verified in this article's verification script — the sign factor in the factorial formula ((−1)^{k−m+m′}, not (−1)^k, for the stated conventions), and the scope of the orthogonality relation (8π² ranges cover SO(3)/integer j; half-integer representations require the SU(2) measure with γ extended to 4π and volume 16π²).

Upstream: Rotation Operators and SU(2) (the operator and the group structure) · Ladder Operators and the Angular Momentum Spectrum ($\hat J_y$'s matrix elements and the phase convention that makes $d$ real) · Angular Momentum Commutation Relations (the algebra behind block-diagonality) · Spherical Harmonics (the integer-$j$ reduction target) · Spin-1/2 and the Pauli Matrices (the exponential identity behind $d^{1/2}$). Downstream: Addition of Angular Momenta and Clebsch-Gordan Coefficients (the series (6) made systematic) · Spherical Tensor Operators and The Wigner-Eckart Theorem (operators that rotate by $D^{k}$) · The Stern-Gerlach Experiment (tilted-analyzer probabilities in the laboratory).