Degeneracy and SU(3) Symmetry of the 3D Oscillator

Elliott's program is alive at scale: modern symmetry-adapted no-core shell model computations organize their gigantic oscillator bases into SU(3) multiplets, cutting basis dimensions by orders of magnitude because real nuclei stubbornly prefer low-deformation multiplets — the 1958 idea running on 2020s supercomputers. Ultracold atoms in optical traps realize the algebra directly: near-isotropic traps show the shell degeneracies, and deliberately tuned anisotropy dials the symmetry breaking of this article's first limiting case in a controlled experiment. And in quantum information, the bilinear construction $\hat{a}_i^\dagger\hat{a}_j$ on a handful of bosonic modes — the same operators, verified the same way — generates the "linear optics" transformations at the heart of photonic computing proposals: SU(3) on three waveguides is a beam-splitter network, and shell $N$ is the $N$-photon sector.

Why should moving a quantum from the $x$-axis to the $z$-axis be a symmetry? Because the isotropic oscillator's axes are interchangeable accountants: every axis charges exactly $\hbar\omega$ per quantum. An operation that debits one account and credits another leaves the books — the total energy — untouched. The rotation group knows only part of this freedom (rotations mix the axes as directions); the full freedom mixes the three ladder operators by any $3\times3$ unitary matrix, and that group is U(3).

The picture to carry: rotations slice each shell into ℓ-blocks and protect the blocks; the extra SU(3) generators are bridges between the blocks, moving states from the $\ell=0$ singlet into the $\ell=2$ quintet without changing the energy. A degeneracy is exactly as large as the set of states its symmetry can reach from one of them — rotations reach $2\ell+1$; SU(3) reaches the whole shell. This is the second panel of a diptych. The first panel is hydrogen: there the bridge operators came packaged as a hidden conserved vector (rank 1), the quantum Runge–Lenz vector of The Laplace–Runge–Lenz Vector in Quantum Mechanics, and the algebra closed into SO(4) — the story of Accidental Degeneracy and the SO(4) Symmetry, told there and only cited here. Classically the two panels share a signature: $1/r$ and $r^2$ are the only central potentials in which every bound orbit closes (Bertrand's theorem, stated here in one line as the mechanics track's result). Closed orbits, conserved extras, degenerate towers — one phenomenon in three languages.

Write the Hamiltonian in the ladder grammar of Ladder Operators and the Algebraic Oscillator, one copy per axis (that article owns the $\hat{a},\hat{a}^\dagger$ machinery; we consume $[\hat{a}_i,\hat{a}_j^\dagger]=\delta_{ij}$, $[\hat{a}_i,\hat{a}_j]=0$ without re-derivation):

where: the index $i$ runs over the Cartesian axes, and the three number operators $\hat{n}_i=\hat{a}_i^\dagger\hat{a}_i$ commute with each other — this is the Cartesian solution of the previous article in operator dress. Now define the object this article is about:

where: for $i\ne j$, $\hat{A}_{ij}$ destroys one quantum on axis $j$ and creates one on axis $i$ — a transfer, not a creation: it maps shell $N$ into shell $N$. The nine $\hat{A}_{ij}$ include the three number operators on the diagonal, and their sum is the total number operator $\hat{N}$, which carries no new information. The off-diagonal six, plus the two independent traceless diagonal combinations, are the news.

Angular momentum must live inside this algebra — rotations are, after all, among the free shuffles. It does, as the antisymmetric part:

where: the three antisymmetric combinations of the off-diagonal bilinears are exactly the components of $\hat{\mathbf{L}}$ — verified matrix-by-matrix against the first-quantized construction $\hat{\mathbf{L}}=\hat{\mathbf{r}}\times\hat{\mathbf{p}}$ of Orbital Angular Momentum Operators (its definition; this article only re-expresses it). The decomposition of the eight su(3) generators into 3 antisymmetric (the $\hat{L}_i$, rank 1 under rotations) plus 5 traceless-symmetric (quadrupole-type, rank 2) is the operator echo of the previous article's $N=2$ worked example, where a symmetric $3\times3$ tensor of states split into trace + traceless parts. Those five quadrupole generators are the bridges: rank-2 tensor operators connect $\ell$ to $\ell\pm2$ (the selection-rule grammar of Spherical Tensor Operators, cited), which is precisely why oscillator shells contain $\ell$ in steps of two.

The two hidden-symmetry stories of this chapter now stand side by side:

where: the comparison is worth a table because the differences teach as much as the parallel. Hydrogen's extra generators form a vector (rank 1: Runge–Lenz, credentials proven in its own serial); the oscillator's form a quadrupole (rank 2: the traceless symmetric bilinears). Hydrogen's degenerate ℓ run in steps of one, the oscillator's in steps of two — each hidden multiplet's ℓ-menu is generated from the top by its bridge operators' rank. Hydrogen's algebra only closes after an energy-dependent rescaling on the bound sector (SO(4) for $E<0$); the oscillator's closes identically, energy sector by energy sector, with no rescaling — the cleaner specimen by far, which is why group-theory courses teach it first.

Example 1 — the machine audit, end to end. Verify the algebra numerically rather than take derivations on faith. (This is this article's verify script, narrated.)

Step 1. Build $\hat{a}_x,\hat{a}_y,\hat{a}_z$ as matrices on a Fock truncation ($n\le6$ per axis, $343$ dimensions). Confirm the canonical algebra on the interior of the truncation: $[\hat{a}_i,\hat{a}_j^\dagger]=\delta_{ij}$, nine of nine (computed).

Step 2. Form $\hat{H}$ and all nine $\hat{A}_{ij}$; confirm $[\hat{H},\hat{A}_{ij}]=0$ exactly — transfers preserve shells even at the truncation edge, because they never change total $N$ (computed, 9/9). Confirm all 81 closure brackets of equation (3) (computed).

Step 3. Confirm the embedding: $\hat{L}_i$ built from bilinears equals $\hat{L}_i$ built from $\hat{\mathbf{r}}\times\hat{\mathbf{p}}$, matrix for matrix (computed, 3/3). Then diagonalize $\hat{L}^2$ on shell blocks: shell 2 yields eigenvalue 0 once and $6\hbar^2$ five times ($\ell=0\oplus2$); shell 3 yields $2\hbar^2$ thrice and $12\hbar^2$ seven times ($\ell=1\oplus3$). The census of the previous article, re-derived by pure algebra — no radial equation, no Laguerre polynomial anywhere in the computation.

Example 2 — the bridge, explicitly. Show one SU(3) generator doing what rotations cannot.

Step 1. Take the $N=2$ s-state $|s\rangle=(|200\rangle+|020\rangle+|002\rangle)/\sqrt3$ (the trace combination identified in the previous article's worked example) and the traceless generator $\hat{Q}=\hat{A}_{xx}-\hat{A}_{yy}$.

Step 2. Compute $|w\rangle=\hat{Q}|s\rangle$: the result is nonzero ($\|w\|=2/\sqrt3$, computed), orthogonal to $|s\rangle$ (overlap $<10^{-12}$), and still an exact $\hat{H}$-eigenstate at $E=\tfrac72\hbar\omega$ (computed). Since the shell's only non-s content is the $\ell=2$ quintet, $|w\rangle$ is pure $\ell=2$: the generator carried a state from the rotational singlet into the quintet without touching the energy.

Step 3. Read the moral: the degeneracy of $\{\ell=0\}\cup\{\ell=2\}$ is not a numerical coincidence to be checked level by level — it is reachability under the symmetry algebra, exhibited here as one matrix-vector multiplication. Answer: $\hat{Q}|s\rangle\perp|s\rangle$, same energy, pure quintet — the "accident" performed live.

• Sakurai & Napolitano §3.7 and problems — the oscillator's degeneracy and its algebra at the source register.
• J.-M. Lévy-Leblond, on global and dynamical symmetries — the cleanest pedagogical treatment of "accidental" degeneracy as hidden symmetry.
• J. P. Elliott, Proc. R. Soc. A 245, 128 & 562 (1958) — nuclear rotations from oscillator SU(3).
• Jauch & Hill, Phys. Rev. 57, 641 (1940) — the original working-out of the oscillator's degeneracy group.

Provenance note: Energy spectrum, degeneracy formula $(N+1)(N+2)/2$, allowed-$\ell$ rule, and the $SU(3)$ generator construction follow Sakurai (Modern QM §3.7 and problems), Shankar Ch. 12–13, and Cohen-Tannoudji Vol. I. A verifier should: (1) confirm $\sum_{\ell=N,N-2,\dots}(2\ell+1)=(N+1)(N+2)/2$ for several $N$; (2) re-derive $[a_i^\dagger a_j,a_k^\dagger a_l]=\delta_{jk}a_i^\dagger a_l-\delta_{il}a_k^\dagger a_j$ from canonical commutators; (3) check the symmetric-tensor dimension $\binom{N+2}{2}$ against the irrep $(N,0)$ of $SU(3)$. The Elliott $SU(3)$ nuclear application is standard (Elliott 1958). Harmonic-oscillator magic numbers $2,8,20,40,70$ and observed magic numbers $2,8,20,28,50,82,126$ are textbook values. — Item (1) is executed in the previous article's verify script (which owns that derivation); items (2) and (3) are executed in this article's verify script (81 closure brackets + 9 conservation commutators as 343-dimensional Fock matrices; dimensions $1,3,6,10,15$ against $\binom{N+2}{2}$), plus the $\hat{L}$-embedding identity, the $\hat{L}^2$ branching spectra on shells 2 and 3, and the ℓ=0→ℓ=2 bridge exhibit. Edition note: the base's opening derivation (equivalence of the Cartesian and spherical degeneracy counts) is owned by The Three-Dimensional Isotropic Harmonic Oscillator per chapter ownership — its third derivation — and is consumed here by citation, not re-proved. Hydrogen's SO(4) machinery is Accidental Degeneracy and the SO(4) Symmetry's throughout; the comparison table quotes its conclusions only.