Separation of Variables in Spherical Coordinates

The angular equation produced by this separation — with its universal spherical-harmonic solutions — reappears essentially unchanged in every problem with spherical or near-spherical symmetry across chemistry and physics. Molecular rotational spectroscopy of diatomic and linear molecules uses precisely the same $\Theta(\theta)\Phi(\phi)$ angular equation (the rigid rotor is mathematically identical to the hydrogen angular problem with $V(r)$ removed entirely), so the microwave and far-infrared rotational transitions used to determine bond lengths in gas-phase molecules trace directly back to this separation. In multi-electron atomic structure theory, the central-field approximation (used in Hartree–Fock and density functional theory for atoms) treats each electron as moving in an effective spherically averaged potential from the nucleus and all other electrons; this approximation is only tractable because the resulting single-particle equations separate exactly the way shown here, producing atomic orbitals labeled by the same $n,\ell,m$ quantum numbers used for hydrogen. Nuclear physics uses the identical angular machinery for nucleons in a spherical shell-model potential, explaining nuclear magic numbers through the same $\ell(\ell+1)$ centrifugal structure. Modern computational chemistry codes that build Gaussian-type or Slater-type atomic basis sets explicitly encode spherical harmonics as the angular factor of every basis function, a direct legacy of this separation. Even scanning-tunneling-microscopy image simulations of atomic orbitals on surfaces, and NMR/EPR calculations involving angular momentum operators, ultimately rest on the same universal angular equation first isolated by separating the Laplacian in spherical coordinates.

Ask the tutor: Can you show step-by-step how the radial equation, once solved for the Coulomb potential, produces the quantized energies $E_n = -13.6\,\text{eV}/n^2$?

Picture the electron not as a point orbiting on a track but as a smeared-out wave $\psi(r,\theta,\phi)$ filling three-dimensional space around the proton. The full time-independent Schrödinger equation is

with $V(r) = -\dfrac{e^{2}}{4\pi\varepsilon_{0}r}$ and $\mu$ the reduced mass of the electron–proton pair. Written out in Cartesian coordinates, $\nabla^{2} = \partial^{2}/\partial x^{2} + \partial^{2}/\partial y^{2} + \partial^{2}/\partial z^{2}$, and $r = \sqrt{x^{2}+y^{2}+z^{2}}$ appears inside a square root buried in $V$. Every attempt to write $\psi = X(x)Y(y)Z(z)$ fails, because the potential refuses to split along $x$, $y$, $z$ individually. The fix is to stop fighting the coordinate system and instead choose one that mirrors the physics. In spherical coordinates, a point is located by its distance from the origin $r\in[0,\infty)$, its polar angle $\theta\in[0,\pi]$ measured from the $z$-axis, and its azimuthal angle $\phi\in[0,2\pi)$ measured around the $z$-axis. The enormous advantage: $V(r) = -e^{2}/4\pi\varepsilon_{0}r$ depends on only one of these three variables. All the angular dependence has been swept entirely out of the potential and now lives purely in the kinetic-energy (Laplacian) operator. That operator, in spherical coordinates, reads

Every angular term carries a prefactor of $1/r^{2}$ — this is the structural signature that makes separation possible, because multiplying through by $r^{2}$ cleanly isolates a radial derivative term from a purely angular operator (recognizable as $-\hat L^{2}/\hbar^{2}$, the squared angular momentum operator in disguise).

The strategy, then, is the standard separation-of-variables playbook applied twice. First split off the radial dependence: guess $\psi(r,\theta,\phi) = R(r)Y(\theta,\phi)$. Substituting into the Schrödinger equation and dividing everything by $R(r)Y(\theta,\phi)$ produces one side that depends only on $r$ and one side that depends only on $(\theta,\phi)$; since $r$ and $(\theta,\phi)$ are independent variables, both sides must equal the same constant. Second, split $Y(\theta,\phi) = \Theta(\theta)\Phi(\phi)$ and repeat the argument — the $\phi$-only piece and the $\theta$-only piece must again share a common constant. Three coupled variables become three uncoupled ordinary differential equations, cascading from the general PDE down to single-variable problems you can solve with power series or known special functions. The constants that appear at each stage are not arbitrary — physical requirements (finiteness, periodicity, single-valuedness of the wavefunction) force them to be quantized. The first separation constant becomes $\ell(\ell+1)$ with $\ell = 0,1,2,\dots$ (the orbital angular momentum quantum number); the second becomes $m^{2}$ with $m = -\ell,\dots,+\ell$ (the magnetic quantum number). These are not put in by hand — they emerge as the only way the separated equations can have physically acceptable solutions.

The starting point is the full time-independent Schrödinger equation for a single particle of reduced mass $\mu$ moving in a central potential $V(r)$:

where:

• $\hbar$ is the reduced Planck constant, $\mu$ the reduced mass of the two-body system,
• $r,\theta,\phi$ are spherical coordinates centered on the nucleus,
• $V(r)$ is any potential depending on distance alone (Coulomb for hydrogen, but the separation works for any central potential),
• $E$ is the total energy eigenvalue,
• $\psi = \psi(r,\theta,\phi)$ is the full three-dimensional wavefunction.

Reading: this equation says the electron's total energy is the sum of a radial kinetic contribution, an angular kinetic contribution (built entirely from $\theta$- and $\phi$-derivatives), and the potential energy — and because $V$ carries no angular dependence, the angular kinetic terms are the only place angles ever appear. Assume a product solution:

where:

• $R(r)$ carries all the radial dependence,
• $Y(\theta,\phi)$ carries all the angular dependence, later itself factored as $\Theta(\theta)\Phi(\phi)$.

Reading: this is an ansatz — a proposed form of solution, justified after the fact by the equation splitting cleanly and by the completeness of the resulting basis of product functions. Substituting and multiplying through by $-2\mu r^{2}/(\hbar^{2}RY)$ produces the separated form:

where:

• the left side depends only on $r$,
• the right side depends only on $\theta$ and $\phi$,
• both sides must therefore equal the same constant, conventionally written $\ell(\ell+1)$.

Reading: an expression that depends purely on $r$ can equal an expression that depends purely on angles, for all values of $r,\theta,\phi$, only if both are equal to one fixed number — this is the entire logical engine of separation of variables. Setting each side equal to $\ell(\ell+1)$ produces two decoupled equations. The radial equation:

where:

• $\ell = 0,1,2,\dots$ is the orbital angular momentum quantum number, forced to be a non-negative integer by the angular boundary conditions,
• $\ell(\ell+1)\hbar^{2}/r^{2}$ acts exactly like an extra repulsive "centrifugal" potential term,
• this single ODE, solved for $V(r) = -e^{2}/4\pi\varepsilon_{0}r$, yields the radial wavefunctions and the Bohr energies.

And the angular equation:

where:

• this equation contains no reference to $V(r)$ or $E$ at all — it is universal for every central-force problem, from the hydrogen atom to a rigid rotor to a particle on a sphere,
• its solutions, the spherical harmonics $Y_{\ell}^{m}(\theta,\phi)$, are shared verbatim across all these systems.

Reading: the entire chemical identity of the potential (Coulomb, harmonic, square well, anything central) survives only in the radial equation; the angular equation is pure geometry, blind to what force is actually acting.

Example 1 — Foundational: verifying the separation constant algebraically for $\ell = 2$. Take the angular equation with the separation constant already fixed at $\ell(\ell+1)$ for $\ell=2$, so $\ell(\ell+1) = 2\times3 = 6$. Confirm directly that a known angular solution satisfies it. Take $m=0$ and the associated Legendre polynomial $P_{2}^{0}(\cos\theta) = \tfrac{1}{2}(3\cos^{2}\theta - 1)$, a valid $\Theta(\theta)$ for $\ell=2, m=0$. Let $x = \cos\theta$, so $\Theta = \tfrac12(3x^{2}-1)$, and recall the polar equation in this substitution (the Legendre equation) reads

Compute $\dfrac{d\Theta}{dx} = 3x$, so $(1-x^{2})\dfrac{d\Theta}{dx} = 3x - 3x^{3}$. Differentiate again:

Now add $\ell(\ell+1)\Theta = 6\times\tfrac12(3x^{2}-1) = 9x^{2}-3$:

The equation is satisfied identically for all $x$, confirming that $\ell=2$ and $\Theta = P_2^0(\cos\theta)$ are mutually consistent, exactly as the separation argument predicts. As a limiting check: setting $\ell=0$ instead gives $\Theta =$ const (since $P_0^0=1$), and indeed $\frac{d}{dx}[(1-x^2)\cdot 0] + 0\cdot 1 = 0$ trivially — the $s$-orbital angular part is simply a constant, as it must be since $s$ orbitals are spherically symmetric.

Example 2 — Applied: reading off the centrifugal term for a $3d$ electron in hydrogen. For a hydrogen $3d$ orbital, $n=3$, $\ell=2$. The radial equation derived above contains the term

acting as an effective repulsive potential added to $V(r)$ (this grouping is standard once the radial equation is rearranged into an effective one-dimensional Schrödinger form for $u(r)=rR(r)$). Evaluate this centrifugal term numerically at $r = 3a_{0}$ (roughly the most probable radius region for $n=3$ states), where $a_0 = 5.29\times10^{-11}\,\text{m}$ is the Bohr radius. With $\ell=2$: $\ell(\ell+1) = 6$.

(using $\mu \approx m_e$ to three figures, since the reduced-mass correction is under 0.06%). Compute the numerator:

Compute $r^2$: $r = 1.587\times10^{-10}\,\text{m}$, so $r^2 = 2.519\times10^{-20}\,\text{m}^2$. Denominator: $2\times9.109\times10^{-31}\times2.519\times10^{-20} = 4.590\times10^{-50}\,\text{kg}\,\cdot\,\text{m}^2$.

This centrifugal contribution (about 9.1 eV at this radius) is comparable in magnitude to the total hydrogen ground-state energy scale (13.6 eV), illustrating why $d$ orbitals ($\ell=2$) are pushed to much larger characteristic radii than $s$ orbitals of the same $n$ — the angular momentum barrier physically excludes the electron from small $r$. As a limiting check, for an $s$ orbital ($\ell=0$) this term vanishes identically at every $r$, consistent with $s$-electron densities having no centrifugal exclusion from the nucleus (and hence the famous nonzero probability of an $s$ electron being found at $r=0$).

Example 3 — Competition-level: showing the separation constant cannot be a non-half-integer-related irrational number by boundary-condition analysis. Suppose (for contradiction, as a rigor check) that the separation constant in the polar equation were not $\ell(\ell+1)$ for integer $\ell$, but instead an arbitrary constant $\lambda$, and ask what value of $\lambda$ is forced by requiring $\Theta(\theta)$ finite at both poles $\theta=0,\pi$ for the $m=0$ case. The Legendre equation with general $\lambda$ is

This is a Sturm–Liouville problem with singular endpoints at $x=\pm1$ (since the coefficient $(1-x^2)$ vanishes there). Attempt a power-series (Frobenius) solution about $x=0$:

Substituting into the ODE and collecting powers of $x^{k}$ produces the recursion

For generic (non-quantized) $\lambda$, this recursion never terminates — the series continues to all orders and, by a standard asymptotic ratio test as $k\to\infty$ (the ratio $a_{k+2}/a_k \to 1$), the series diverges at $x=\pm1$, i.e. exactly at the poles $\theta = 0,\pi$. A divergent $\Theta$ at the poles is physically unacceptable (the wavefunction must remain finite everywhere on the sphere, including the poles, since these are perfectly ordinary points in three-dimensional space — nothing physically singular happens there). The only way to avoid divergence is for the series to terminate after a finite number of terms, turning it into a polynomial. Setting the numerator to zero at $k=\ell$ (some non-negative integer) gives

So the demand for a finite, physically sensible wavefunction at the poles — not an ad hoc assumption — is exactly what forces the separation constant to take the discrete values $0, 2, 6, 12, 20,\dots$ (i.e. $\ell(\ell+1)$ for integer $\ell$) rather than any continuous or irrational value. This is the rigorous origin of orbital angular momentum quantization purely from the separation-of-variables machinery, before any operator algebra with $\hat L^2$ is even introduced.

The separation-of-variables method for the Laplacian in spherical coordinates predates quantum mechanics by more than a century, developed for problems in gravitation, electrostatics, and heat conduction. Its formal quantum-mechanical application to the hydrogen atom was carried out by Erwin Schrödinger himself in his 1926 papers, drawing on classical mathematical physics techniques already well established by Adrien-Marie Legendre (whose polynomials solve the polar equation) and Pierre-Simon Laplace (whose equation in spherical coordinates is the direct ancestor of the angular equation here).