Ladder Operators and the Angular Momentum Spectrum

Every two-level quantum device is a $j=\tfrac12$ ladder; the master formula is why microwave and laser pulses steer qubits between rungs with calibrated amplitudes. In MRI and NMR the proton's two-rung ladder is driven at its Larmor frequency, with signal strengths traceable to matrix elements of $\hat J_\pm$. Atomic clocks and quantum sensors park atoms in stretched states $|j,\pm j\rangle$ because — as Example 2 shows — those saturate the uncertainty bound. And every quantum-chemistry code hard-wires the $\sqrt{j(j+1)-m(m\pm1)}$ factors into its Clebsch–Gordan libraries. The 1925 square root never retired; it moved into firmware.

Picture the eigenstates of $\hat J_z$ as rungs on a ladder. Rungs of one ladder share the same $\hat{\mathbf J}^2$ — the same total vector length $\sqrt{j(j+1)}\,\hbar$ — and differ only in the projection $m\hbar$, spaced exactly $\hbar$ apart. $\hat J_+$ climbs one rung, $\hat J_-$ descends one, and neither ever changes ladders. The spectrum problem reduces to two questions: how long can a ladder be, and which lengths are allowed?

The second picture explains why the ladder stops. The vector has squared length $j(j+1)\hbar^2$, but its largest projection is only $j\hbar$ — strictly shorter than the vector. Even on the top rung, $\langle\hat J_x^2+\hat J_y^2\rangle=\hbar^2[j(j+1)-j^2]=j\hbar^2>0$: some length always lives in the transverse plane, and angular momentum never points exactly along the measured axis. Each rung is not an arrow but a cone; climbing past $m=j$ would demand a projection longer than the vector — a negative transverse spread, a negative norm. The algebra refuses.

Hold both pictures and the derivation is three sentences: the commutators move $m$ in unit steps without changing $j$; positivity caps $|m|$; and demanding that the capped ladder's two annihilated ends belong to the same ladder forces $2j$ to be a whole number. The rest is bookkeeping — bookkeeping that runs atomic physics.

Everything here is manufactured from one definition — fuse the two Hermitian transverse components into a complex pair, the move that turned $\hat x,\hat p$ into $\hat a,\hat a^\dagger$:

where: $\hat J_x,\hat J_y$ are the Hermitian Cartesian components and $\hat J_\pm$ are non-Hermitian mutual adjoints. They are not observables — nothing measures $\hat J_+$ — they are machines that transform states. Notice there is no system-specific input here: any triple of operators obeying the angular-momentum algebra of the previous article gets a ladder pair for free. The three working commutators follow in two lines each — internalize them as verbs: "step," "feedback," "accounting."

Because $[\hat{\mathbf J}^2,\hat J_z]=0$, simultaneous eigenstates $|\lambda,m\rangle$ carry both labels — Angular Momentum Commutation Relations showed these two are all you get:

where: $\lambda$ is dimensionless (that it equals $j(j+1)$ is the theorem to come) and $m$ counts the $z$-projection in units of $\hbar$. The stepping relation gives $\hat J_z\big(\hat J_\pm|\lambda,m\rangle\big) = \hbar(m\pm1)\,\hat J_\pm|\lambda,m\rangle$: the state $\hat J_\pm|\lambda,m\rangle$, if it is not zero, sits one rung up or down. That italic clause is where all the physics hides.

Pause on what was not used: no coordinates, no single-valuedness, no Hamiltonian. Where the square well was quantized by boundary conditions, the "boundary" here is positivity of the inner product itself — so one table of numbers governs an electron's spin, a molecule's tumbling, and a nucleus's ground state.

Termination said where the ladder ends; the accounting identity also prices every step. Write $\hat J_+|j,m\rangle = c_{jm}^{+}\,|j,m{+}1\rangle$ and compute the norm of the left side:

where: the middle step substitutes eigenvalues into $\hat J_-\hat J_+=\hat{\mathbf J}^2-\hat J_z^2-\hbar\hat J_z$. The norm fixes $|c_{jm}^{+}|$ — and only the modulus; the phase is genuinely undetermined.

For each $j$ the master formula fills in a matrix representation of dimension $2j+1$. Order the basis $|j,j\rangle,\dots,|j,-j\rangle$: $\hat J_z$ is diagonal, $\hat J_+$ has one nonzero super-diagonal, and $\hat J_x,\hat J_y$ follow by inverting the defining split. For $j=\tfrac12$ the two-rung ladder yields, in one line, $\hat S_i=\tfrac\hbar2\sigma_i$ with $\hat{\mathbf S}^2=\tfrac34\hbar^2\mathbb 1$ — the Pauli structure, derived before anyone mentions spin; its anatomy is owned by Spin-1/2 and the Pauli Matrices. For $j=1$ the three-rung ladder gives matrices worth building once by hand; the only inputs are the two step coefficients, both $\hbar\sqrt2$:

where: both off-diagonal strengths are $\hbar\sqrt2$ (steps into and out of the middle rung), and $\hat{\mathbf J}^2=2\hbar^2\mathbb 1$ as $j(j+1)$ demands. A worthwhile check: the eigenvalues of $\hat J_x$ are $+\hbar,0,-\hbar$ — the same rung values as $\hat J_z$. Of course they are: no axis is special, and the ladder you would build along $x$ is the same ladder.

The derivation allowed $j=\tfrac12,\tfrac32,\dots$, and nature uses them — but never for orbital angular momentum. That eviction is not decided by this article's algebra: $\hat L_z=-i\hbar\,\partial_\phi$ acts on wavefunctions of a spatial angle, and demanding $e^{im\phi}$ take one value per point of space forces $m$ — hence $\ell$ — into the integers. The single-valuedness argument and why it binds $\hat{\mathbf L}$ but not spin are owned by Eigenfunctions of L_z and the Azimuthal Quantization. The algebra supplies the shape of the pigeonholes; which ones a system occupies is extra physics.

You solved the harmonic oscillator by ladder once — Ladder Operators and the Algebraic Oscillator owns that spectrum, and nothing from it is re-derived here. But the side-by-side comparison teaches this article's deepest lesson:

The constant commutator gives the oscillator an endless staircase; the operator-valued one gives angular momentum a self-limiting ladder. Whenever you meet a new algebra, ask first: is the commutator of the ladders a constant or an operator? That decides whether the spectrum is a half-line or a box.

Example 1 — build the $j=1$ matrices and check an eigenvalue claim. Task: construct $\hat J_x$ for $j=1$ from the master formula and find its eigenvalues.

Step 1. The nonzero raising elements are $\langle 1,1|\hat J_+|1,0\rangle$ and $\langle 1,0|\hat J_+|1,-1\rangle$. Both coefficients evaluate to $\hbar\sqrt{1\cdot2-0\cdot1}=\hbar\sqrt{2}$ and $\hbar\sqrt{1\cdot2-(-1)\cdot0}=\hbar\sqrt{2}$.

Step 2. Assemble $\hat J_x=(\hat J_++\hat J_-)/2$: each $\hbar\sqrt2$ splits symmetrically into $\hbar/\sqrt2 \approx 0.707\,\hbar$, giving the tridiagonal matrix displayed above.

Step 3. Diagonalize (or trust isotropy): the eigenvalues are $\boldsymbol{+\hbar,\ 0,\ -\hbar}$ — the same rungs as $\hat J_z$, as rotational symmetry demands; numerically $\pm1.000\,\hbar$ and $0$ to machine precision.

Step 4. Sanity check the Casimir: $\hat J_x^2+\hat J_y^2+\hat J_z^2 = 2\hbar^2\,\mathbb 1 = 1(1{+}1)\hbar^2\,\mathbb 1$. The master formula plus two square roots reproduces the entire $j=1$ triple.

Example 2 — the cone of a stretched state. Task: a nucleus sits in the stretched state $|j,m\rangle=|2,2\rangle$. How long is its angular-momentum vector, how far does it tilt from the $z$-axis, and what are the transverse uncertainties?

Step 1. Length: $|\mathbf J| = \sqrt{j(j+1)}\,\hbar = \sqrt6\,\hbar \approx \boldsymbol{2.449\,\hbar}$, while the projection is only $m\hbar = 2\hbar$.

Step 2. Tilt: $\cos\theta = m/\sqrt{j(j+1)} = 2/\sqrt6$, so $\theta \approx \boldsymbol{35.26^\circ}$. Even "maximally aligned," the vector rides a cone a third of the way to the equator.

Step 3. Transverse spread: $\langle\hat J_x^2\rangle = \langle\hat J_y^2\rangle = \tfrac{\hbar^2}{2}\big[j(j+1)-m^2\big] = \tfrac{\hbar^2}{2}(6-4) = \hbar^2$, so $\sigma_{J_x}=\sigma_{J_y}=\boldsymbol{\hbar}$ exactly (with $\langle\hat J_x\rangle=\langle\hat J_y\rangle=0$).

Step 4. Check against the uncertainty relation of the previous article: $\sigma_{J_x}\sigma_{J_y} = \hbar^2 = \tfrac\hbar2|\langle\hat J_z\rangle| = \tfrac\hbar2\cdot2\hbar$. The stretched state saturates the bound exactly — it is the most classical orientation state the algebra allows, and still it tilts.

References. Sakurai & Napolitano, Modern Quantum Mechanics, 3rd ed., §3.5 · Griffiths, Introduction to Quantum Mechanics, 3rd ed., §4.3 · Shankar, Principles of Quantum Mechanics, 2nd ed., Ch. 12 · Cohen-Tannoudji, Diu & Laloë, Quantum Mechanics, Vol. I, Ch. VI · Townsend, A Modern Approach to Quantum Mechanics, Ch. 3.

Provenance. All commutators and the spectrum can be recomputed from $[J_i,J_j]=i\hbar\epsilon_{ijk}J_k$ as in the derivations; verify $\mathbf{J}^2=J_\mp J_\pm+J_z^2\pm\hbar J_z$ and the coefficient $\sqrt{j(j+1)-m(m\pm1)}$ against Sakurai §3.5, Griffiths §4.3, Shankar Ch. 12, and Cohen-Tannoudji Ch. VI. The spin-1/2 matrices reduce to the standard Pauli matrices (Condon–Shortley phase). No empirical constants are used beyond $\hbar$; all numbers ($j(j+1)$, $2j+1$, $\tfrac34\hbar^2$ for spin-1/2) are exact algebraic values.

Upstream: Angular Momentum Commutation Relations (the defining algebra and the Casimir, consumed whole) · Orbital Angular Momentum Operators (the concrete $\hat{\mathbf L}$ realization) · Ladder Operators and the Algebraic Oscillator (the constant-commutator cousin). Downstream: Spherical Harmonics (the integer ladders realized on the sphere) · Eigenfunctions of L_z and the Azimuthal Quantization (who evicts the half-integers from orbits) · Spin Angular Momentum and Spin-1/2 and the Pauli Matrices (who houses them) · Clebsch-Gordan Coefficients (where the square roots multiply).