Separation of Variables in Multidimensional Problems

A single hydrogen atom moving in three dimensions faces a Schrödinger equation. It has one wavefunction and three spatial coordinates tangled together. Yet chemists routinely draw energy levels labeled by three separate quantum numbers. This makes it look as if the problem had split into three independent one-dimensional puzzles. It has. The technique responsible is separation of variables. It is the single mathematical trick that converts almost every solvable model in quantum chemistry. This includes the particle in a 3D box, the harmonic oscillator, the rigid rotor, and the hydrogen atom. Separation of variables transforms an intractable partial differential equation into a set of ordinary differential equations. These can be solved one coordinate at a time. This matters because real molecules live in three-dimensional space. Their vibrational, rotational, and electronic degrees of freedom are all entangled in principle. Without separability, the Schrödinger equation for even a diatomic molecule would need more. It would require solving a coupled multidimensional PDE with no closed-form structure. Separation of variables is the mathematical statement of a physical idea. When a Hamiltonian has no cross-terms coupling coordinates, motions along each coordinate are dynamically independent. The total wavefunction then factors into a product of one-dimensional wavefunctions. Their energies simply add. The big picture is that separability is not automatic. It is a structural property of the Hamiltonian. It depends on the potential energy function being additively decomposable in an appropriate coordinate system. This decomposition must be additive, not multiplicative or otherwise. When that condition holds, the total energy eigenvalue equation collapses into independent eigenvalue equations. Each is governed by its own separation constant. The degeneracy patterns of the full problem emerge directly from this structure. They depend on how many ways the separation constants can combine to give the same total energy. This article develops the general machinery. It shows how the ansatz $\Psi = X(x)Y(y)Z(z)$ (or its curvilinear analogues) is justified. This justification comes from the structure of the Laplacian and the potential. It also shows how separation constants arise and are interpreted. Finally, it shows how the resulting factored energies and degeneracies are read directly off the separated equations. Worked examples move from the 3D box through an anisotropic 3D harmonic oscillator. They continue to a general asymmetric box with three unequal edge lengths. Together these examples build the toolkit underlying the particle on a ring and the particle on a sphere.

Key takeaways

Separation of variables works when the Hamiltonian is a sum of independent one-coordinate pieces: $\hat H = \hat H_x + \hat H_y + \hat H_z$ , with no cross terms in the potential.
The ansatz $\Psi(x,y,z) = X(x)Y(y)Z(z)$ turns the PDE into ODEs, each equal to a separation constant ( $E_x$ , $E_y$ , $E_z$ ), because a function of $x$ alone can only equal a function of $y,z$ alone if both equal the same constant.
The total energy is the sum of the 1D eigenvalues, $E = E_x + E_y + E_z$ , and the total wavefunction is the product of the 1D eigenfunctions.
Degeneracy arises when different combinations of separated quantum numbers give the same total energy — this is purely arithmetic once the 1D spectra are known.
The technique generalizes to any coordinate system (Cartesian, spherical, cylindrical) as long as the Laplacian and potential are separable in that system — this is exactly what enables the particle on a sphere and hydrogen atom to be solved.
Separability is a property of the potential's functional form, not of the number of dimensions — a 3D problem with a coupling term like $xy$ generally does not separate in Cartesian coordinates.
Key idea: if the Hamiltonian splits into additive, non-interacting pieces along each coordinate, the Schrödinger equation splits into independent 1D eigenvalue problems whose energies add and whose eigenfunctions multiply.

Where this lives today

Separation of variables is not a historical curiosity — it is the working engine behind how chemists actually think about molecular energy levels. Vibrational spectroscopy relies on the normal-mode decomposition of a molecule's potential energy surface, in which the (approximately harmonic, mass-weighted) vibrational Hamiltonian is diagonalized into 3N−6 (or 3N−5 for linear molecules) independent normal-mode oscillators — an enormously more elaborate version of the coordinate-rotation trick in Example 3, applied to the full dimensionality of a polyatomic molecule. Infrared and Raman selection rules, group-theoretical mode classification, and force-field fitting all presuppose that the potential can be made separable (at least approximately, ignoring anharmonic coupling) in the right set of coordinates. In electronic structure theory, the separability logic underlies the orbital approximation itself: the exact many-electron Schrödinger equation does not separate because of electron-electron repulsion terms like $1/r_{12}$ , which couple electron coordinates exactly the way the $xy$ term coupled Example 3. The entire machinery of Hartree-Fock and density functional theory can be understood as strategies for approximately restoring separability — replacing the true, coupled electron-electron interaction with an effective, separable mean-field potential so that the many-electron problem again factors into one-electron orbital equations solvable one coordinate (one electron) at a time. In rotational spectroscopy and molecular quantum mechanics more broadly, central potentials (the hydrogen atom, the rigid rotor, any spherically symmetric model potential used in scattering theory) are solved by separating the radial and angular parts in spherical coordinates, which is exactly the route toward the particle on a ring and the particle on a sphere — themselves the direct ancestors of the quantum numbers $\ell$ and $m_\ell$ that organize atomic orbitals, molecular rotational spectra, and even nuclear shell structure. Modern computational chemistry packages, when they diagonalize a vibrational Hessian into normal modes or expand a molecular orbital in spherical harmonics, are implementing this same separation-of-variables logic numerically, coordinate by coordinate, exactly as derived by hand in this article.

Ask the tutor: Can you show me step by step how the hydrogen atom's Schrödinger equation separates into radial and angular parts in spherical coordinates?

The physical picture — thorough prose

Consider the time-independent Schrödinger equation for a single particle of mass $m$ moving in three dimensions under a potential $V(x,y,z)$ :

-\frac{\hbar^2}{2m}\left(\frac{\partial^2\Psi}{\partial x^2} + \frac{\partial^2\Psi}{\partial y^2} + \frac{\partial^2\Psi}{\partial z^2}\right) + V(x,y,z)\,\Psi = E\,\Psi

This is a linear partial differential equation in three independent variables. In general such an equation is formidable — the derivatives with respect to $x$ , $y$ , and $z$ are tied together by the single scalar $\Psi(x,y,z)$ , and there is no reason to expect the solution to have any simple structure. But chemists know that for the particle in a 3D box, the harmonic oscillator, and the hydrogen atom, the solutions do have simple structure: they factor into products of one-dimensional functions. The reason is not a coincidence — it is a direct consequence of the potential energy function's shape. The kinetic energy operator, the Laplacian, is already a sum of independent second derivatives — one per coordinate — with no cross derivatives like $\partial^2/\partial x\,\partial y$ . The only way cross-coupling between $x$ , $y$ , and $z$ can enter the Hamiltonian is through the potential $V$ . If $V$ happens to be additively separable,

V(x,y,z) = V_x(x) + V_y(y) + V_z(z)

then the entire Hamiltonian breaks into three independent pieces:

\hat H = \hat H_x + \hat H_y + \hat H_z, \qquad \hat H_q = -\frac{\hbar^2}{2m}\frac{d^2}{dq^2} + V_q(q)

for $q = x, y, z$ . Physically, this means the force along $x$ depends only on $x$ (not on where the particle is in $y$ or $z$ ), and likewise for $y$ and $z$ . There is no mechanism in the physics for motion in one direction to influence motion in another. This is why the box (where $V_x, V_y, V_z$ are independent infinite walls), the isotropic or anisotropic harmonic oscillator (where $V = \frac12 k_x x^2 + \frac12 k_y y^2 + \frac12 k_z z^2$ ), and the free particle (where $V=0$ everywhere, trivially additive) all separate — while a potential like $V = \lambda xyz$ or $V = -e^2/r$ written in Cartesian coordinates does not separate in $(x,y,z)$ , because $r = \sqrt{x^2+y^2+z^2}$ mixes all three coordinates non-additively. That last example is precisely why the hydrogen atom is solved in spherical coordinates instead — the Coulomb potential $-e^2/(4\pi\epsilon_0 r)$ is a function of $r$ alone, and in spherical coordinates the angular and radial parts of the Laplacian can be peeled apart, which is the deep reason the spherical harmonics appear universally in central-potential problems. Once the Hamiltonian is additively separable, the product ansatz $\Psi(x,y,z) = X(x)Y(y)Z(z)$ converts the PDE into three ordinary differential equations, each involving only one variable, tied together by separation constants that must sum to the total energy. The logic (worked in full below) hinges on a simple but powerful fact: if a function of $x$ alone equals a negative of a sum of functions of $y$ and $z$ alone, for all values of $x$ , $y$ , $z$ independently, then every one of those functions must in fact be a constant. This is the mathematical seed from which quantized 1D sub-problems — and the additive energy-level structure so central to spectroscopy — grow.

Figure

Separable vs. non-separable potentials in the (x,y) plane

Figure

From 3D Hamiltonian to three 1D eigenvalue problems

The core ideas — governing equations

The separability condition. For a Hamiltonian in coordinates $(q_1, q_2, q_3)$ with kinetic energy given by a Laplacian with no cross-derivative terms,

\hat H = -\frac{\hbar^2}{2m}\nabla^2 + V(q_1,q_2,q_3)

separation of variables is possible in Cartesian coordinates whenever

V(x,y,z) = V_x(x) + V_y(y) + V_z(z)

where: $V_x, V_y, V_z$ are one-variable functions; there is no cross term such as $xy$ , $xz$ , $r$ -dependence mixing all three, etc. Reading: additivity of the potential is the structural requirement — the Laplacian is always "pre-separated," so only $V$ can spoil it.

The product ansatz and the separated ODEs. Substituting $\Psi = X(x)Y(y)Z(z)$ into the Schrödinger equation and dividing through by $XYZ$ yields

\left[-\frac{\hbar^2}{2mX}\frac{d^2X}{dx^2} + V_x(x)\right] + \left[-\frac{\hbar^2}{2mY}\frac{d^2Y}{dy^2} + V_y(y)\right] + \left[-\frac{\hbar^2}{2mZ}\frac{d^2Z}{dz^2} + V_z(z)\right] = E

where: each bracket depends on only one variable, and their sum is the constant $E$ . Reading: three "independent-variable" functions summing to a constant force each bracket individually to be constant — this is the separation-constant argument spelled out fully below.

The separated eigenvalue equations and total energy.

-\frac{\hbar^2}{2m}\frac{d^2X}{dx^2} + V_x(x)X = E_x X, \qquad E = E_x + E_y + E_z

(and analogously for $Y$ , $Z$ ) where: $E_x, E_y, E_z$ are separation constants, each the eigenvalue of a genuine 1D Schrödinger equation; $E$ is the total energy. Reading: the multidimensional eigenvalue problem has been converted, exactly, into three independent 1D eigenvalue problems solved by whatever technique applies to each 1D potential (box, oscillator, etc.).

Total wavefunction and normalization.

\Psi_{n_x n_y n_z}(x,y,z) = X_{n_x}(x)\,Y_{n_y}(y)\,Z_{n_z}(z)

where: $n_x, n_y, n_z$ are the quantum numbers of each 1D sub-problem. Reading: because the 1D eigenfunctions are each individually normalized ( $\int|X_{n_x}|^2dx = 1$ , etc.), the product is automatically normalized over the full 3D domain, since the volume integral factors into a product of three 1D integrals.

Degeneracy from the separated spectrum. If the $E_{n_x}, E_{n_y}, E_{n_z}$ formulas admit different quantum-number triples $(n_x,n_y,n_z) \ne (n_x',n_y',n_z')$ giving the same $E$ , those states are degenerate — this is precisely the mechanism explored for the 3D box, where equal edge lengths make certain permutations of quantum numbers accidentally coincide in energy.

Derivation AWhy the separation constant must be a true constant

Goal: show rigorously that when a function of $x$ alone plus a function of $(y,z)$ alone sum to a constant $E$ , each piece must separately equal a constant — the logical heart of the method.

Step 1. Set up the divided equation. Starting from

-\frac{\hbar^2}{2m}\left(Y Z\,X'' + XZ\,Y'' + XY\,Z''\right) + \left[V_x(x)+V_y(y)+V_z(z)\right]XYZ = E\,XYZ

divide every term by $\Psi = XYZ$ (valid wherever $\Psi \ne 0$ ):

\frac{-\frac{\hbar^2}{2m}X''}{X} + V_x(x) \;+\; \frac{-\frac{\hbar^2}{2m}Y''}{Y} + V_y(y) \;+\; \frac{-\frac{\hbar^2}{2m}Z''}{Z} + V_z(z) = E

Define $f(x) \equiv -\frac{\hbar^2}{2mX}X'' + V_x(x)$ and similarly $g(y)$ , $h(z)$ . The equation reads

f(x) + g(y) + h(z) = E

Step 2. Isolate the single-variable piece. Rearrange to

f(x) = E - g(y) - h(z)

The left side depends on $x$ only. The right side, as written, depends on $y$ and $z$ only — crucially, it does not depend on $x$ at all, because $E$ is the fixed total energy and $g,h$ never contain $x$ .

Step 3. Vary $y$ and $z$ while holding $x$ fixed. Pick any fixed value $x_0$ . Then $f(x_0)$ is some fixed number. But the equation says this fixed number equals $E - g(y) - h(z)$ for every choice of $y$ and $z$ — including choices that change $g(y)+h(z)$ . The only way a quantity that varies with $y,z$ can be forced to equal a fixed number for all $y,z$ is if it, too, does not actually vary — i.e., $g(y) + h(z)$ must be a constant, call it $C_{yz}$ .

Step 4. Conclude $f(x)$ is constant. Since $g(y)+h(z) = C_{yz}$ for all $y,z$ , Step 2 gives $f(x) = E - C_{yz} \equiv E_x$ , a constant independent of $x$ . So

f(x) = E_x \quad\text{(a constant, for all } x\text{)}

This is exactly the statement $-\frac{\hbar^2}{2m}X'' + V_x(x)X = E_x X$ — a genuine 1D Schrödinger equation.

Step 5. Repeat for $y$ and $z$ . By the identical argument applied to $g(y) = C_{yz} - h(z)$ (now varying $z$ while $y$ fixed, then noting $h(z)$ must itself be constant), one obtains $g(y) = E_y$ and $h(z) = E_z$ as separate constants with

E_x + E_y + E_z = E

Each separation constant is fixed by requiring the corresponding 1D function ( $X$ , $Y$ , or $Z$ ) to be well-behaved (finite, single-valued, matching boundary conditions) — which is exactly what quantizes $E_x, E_y, E_z$ individually, exactly as in any 1D problem.

Go deeper · Derivation B · 1 minDegeneracy counting from separated quantum numbers: the 3D box as template

Derivation BDegeneracy counting from separated quantum numbers: the 3D box as template

Goal: derive the energy formula and degeneracy structure for a particle in a 3D box directly from separation of variables, to show how "accidental" degeneracies are just arithmetic coincidences among the separated spectra.

Step 1. Confirm separability. The 3D box potential is

V(x,y,z) = \begin{cases} 0 & 0 < x < a,\ 0 < y < b,\ 0 < z < c \\ \infty & \text{otherwise}\end{cases}

Inside the box this is $V = 0 = V_x(x) + V_y(y) + V_z(z)$ with each piece identically zero — trivially additive. Outside, the wavefunction must vanish (infinite wall), which is equivalent to each factor $X, Y, Z$ separately vanishing at its own walls. So the box separates into three independent infinite-well problems.

Step 2. Solve each 1D piece. Each factor solves the particle-in-a-box equation on its own interval, e.g.

-\frac{\hbar^2}{2m}X'' = E_x X, \qquad X(0)=X(a)=0

giving $X_{n_x}(x) = \sqrt{2/a}\sin(n_x\pi x/a)$ with

E_{n_x} = \frac{n_x^2 h^2}{8ma^2}, \qquad n_x = 1,2,3,\dots

and analogous formulas with $(n_y, b)$ and $(n_z, c)$ .

Step 3. Assemble the total energy. By the separation-constant result,

E_{n_x n_y n_z} = \frac{h^2}{8m}\left(\frac{n_x^2}{a^2} + \frac{n_y^2}{b^2} + \frac{n_z^2}{c^2}\right)

and the full wavefunction is the triple product

\Psi_{n_x n_y n_z}(x,y,z) = \sqrt{\frac{8}{abc}}\;\sin\!\left(\frac{n_x\pi x}{a}\right)\sin\!\left(\frac{n_y\pi y}{b}\right)\sin\!\left(\frac{n_z\pi z}{c}\right)

Step 4. Read off degeneracy for the cube ( $a=b=c$ ). When the box is a cube, $E \propto n_x^2+n_y^2+n_z^2$ , so any permutation of a quantum-number triple gives the same energy. For instance $(1,1,2)$ , $(1,2,1)$ , $(2,1,1)$ all give $n_x^2+n_y^2+n_z^2 = 6$ : this is a genuine, symmetry-required 3-fold degeneracy. But $(3,0,0)$ -type combinations are excluded since $n=0$ is not allowed; a subtler case is $(1,2,2)^2$ -sum $=1+4+4=9$ matching $(0,\dots)$ — no valid alternative, so that level stays non-degenerate at the level of small integers, but larger boxes generically show extra "accidental" degeneracies when different (non-permutation) triples happen to give the same sum of squares (a number-theoretic coincidence, e.g. $(1,5,5)$ and $(1,1,7)$ both giving sum $51$ — sometimes called accidental degeneracy since it is not required by the geometric symmetry of the permutation group).

Step 5. Break degeneracy by desymmetrizing edges. If $a \ne b \ne c$ , the energy formula loses its permutation symmetry, and states that were degenerate for the cube split apart — this is exactly the content of degeneracy lifting when box symmetry is broken, and it is reproduced quantitatively in the worked example below with a fully asymmetric box.

Worked examples

Example 1 — Foundational: verifying the ground and first excited states of a 3D box with unequal edges. Take a box with $a = 2.00\ \text{Å}$ , $b = 3.00\ \text{Å}$ , $c = 4.00\ \text{Å}$ (deliberately unequal, asymmetric, to stress-test the formula without any symmetry hiding an error), containing an electron ( $m = 9.109\times10^{-31}\ \text{kg}$ ). The separated energy formula is

E_{n_x n_y n_z} = \frac{h^2}{8m}\left(\frac{n_x^2}{a^2} + \frac{n_y^2}{b^2} + \frac{n_z^2}{c^2}\right)

Compute the prefactor:

\frac{h^2}{8m} = \frac{(6.626\times10^{-34})^2}{8\times 9.109\times10^{-31}} = \frac{4.390\times10^{-67}}{7.287\times10^{-30}} = 6.025\times10^{-38}\ \text{J}\,\cdot\,\text{m}^2

Convert edge lengths to meters: $a = 2.00\times10^{-10}\,\text{m}$ , $b=3.00\times10^{-10}\,\text{m}$ , $c=4.00\times10^{-10}\,\text{m}$ , so

\frac{1}{a^2} = 2.500\times10^{19}\,\text{m}^{-2}, \quad \frac{1}{b^2} = 1.111\times10^{19}\,\text{m}^{-2}, \quad \frac{1}{c^2} = 6.250\times10^{18}\,\text{m}^{-2}

Ground state $(1,1,1)$ :

E_{111} = 6.025\times10^{-38}\left(2.500\times10^{19} + 1.111\times10^{19} + 6.250\times10^{18}\right)

= 6.025\times10^{-38}\times 4.236\times10^{19} = 2.552\times10^{-18}\ \text{J} = 15.93\ \text{eV}

Lowest excitation choices. Because $a<b<c$ , raising $n_x$ costs the most energy per step (smallest edge → largest $1/a^2$ ), so the lowest excited state should bump $n_z$ (largest edge, cheapest step) from 1 to 2, i.e., state $(1,1,2)$ :

E_{112} = 6.025\times10^{-38}\left(2.500\times10^{19}+1.111\times10^{19}+4\times6.250\times10^{18}\right)

= 6.025\times10^{-38}\times(2.500+1.111+2.500)\times10^{19} = 6.025\times10^{-38}\times 6.111\times10^{19}

= 3.682\times10^{-18}\ \text{J} = 22.98\ \text{eV}

Compare with bumping $n_y$ instead, state $(1,2,1)$ :

E_{121} = 6.025\times10^{-38}\left(2.500 + 4\times1.111 + 0.625\right)\times10^{19} = 6.025\times10^{-38}\times 7.569\times10^{19} = 4.560\times10^{-18}\ \text{J} = 28.46\ \text{eV}

and bumping $n_x$ , state $(2,1,1)$ :

E_{211} = 6.025\times10^{-38}\left(4\times2.500+1.111+0.625\right)\times10^{19} = 6.025\times10^{-38}\times 11.74\times10^{19} = 7.070\times10^{-18}\ \text{J} = 44.13\ \text{eV}

As expected from the separated-energy formula, $E_{112} < E_{121} < E_{211}$ : the first excited state is $(1,1,2)$ , confirming that the smallest quantum-number increment along the longest edge costs the least energy — a direct, quantitative consequence of separability, since each direction's contribution genuinely adds independently rather than interacting.

Limiting-case check: setting $a=b=c$ recovers the familiar cubic-box formula $E = (h^2/8ma^2)(n_x^2+n_y^2+n_z^2)$ , and then $(1,1,2)$ , $(1,2,1)$ , $(2,1,1)$ become exactly degenerate — consistent with the general degeneracy argument above.

Example 2 — Applied: an anisotropic 3D harmonic oscillator (a simple model for a bent triatomic's local vibration). Consider a particle in a 3D harmonic potential with three different force constants, modeling, say, an anisotropic local vibrational environment:

V(x,y,z) = \tfrac12 k_x x^2 + \tfrac12 k_y y^2 + \tfrac12 k_z z^2

This is additively separable by inspection. Each direction is an independent 1D harmonic oscillator with its own frequency $\omega_q = \sqrt{k_q/m}$ , so

E_{n_x n_y n_z} = \hbar\omega_x\left(n_x+\tfrac12\right) + \hbar\omega_y\left(n_y+\tfrac12\right) + \hbar\omega_z\left(n_z+\tfrac12\right)

Take representative (asymmetric, non-round) force constants $k_x = 350\ \text{N/m}$ , $k_y = 500\ \text{N/m}$ , $k_z = 620\ \text{N/m}$ , with reduced mass $\mu = 1.20\times10^{-26}\ \text{kg}$ (roughly a light triatomic reduced mass).

\omega_x = \sqrt{\frac{350}{1.20\times10^{-26}}} = \sqrt{2.917\times10^{28}} = 1.708\times10^{14}\ \text{s}^{-1}

\omega_y = \sqrt{\frac{500}{1.20\times10^{-26}}} = \sqrt{4.167\times10^{28}} = 2.041\times10^{14}\ \text{s}^{-1}

\omega_z = \sqrt{\frac{620}{1.20\times10^{-26}}} = \sqrt{5.167\times10^{28}} = 2.273\times10^{14}\ \text{s}^{-1}

Zero-point energy (all $n=0$ ):

E_{000} = \tfrac12\hbar(\omega_x+\omega_y+\omega_z) = \tfrac12(1.0546\times10^{-34})(1.708+2.041+2.273)\times10^{14}

= \tfrac12(1.0546\times10^{-34})(6.022\times10^{14}) = \tfrac12(6.350\times10^{-20}) = 3.175\times10^{-20}\ \text{J} = 0.1982\ \text{eV}

Now find the energy of the state $(n_x,n_y,n_z) = (1,0,2)$ :

E_{102} = \hbar\omega_x(1.5) + \hbar\omega_y(0.5) + \hbar\omega_z(2.5)

= (1.0546\times10^{-34})\left[1.708\times10^{14}(1.5) + 2.041\times10^{14}(0.5) + 2.273\times10^{14}(2.5)\right]

= (1.0546\times10^{-34})\left[2.562\times10^{14} + 1.0205\times10^{14} + 5.6825\times10^{14}\right]

= (1.0546\times10^{-34})(9.265\times10^{14}) = 9.771\times10^{-20}\ \text{J} = 0.6100\ \text{eV}

This example makes the anisotropy explicit: since $\omega_x \ne \omega_y \ne \omega_z$ , no accidental degeneracies occur (unlike the isotropic 3D oscillator, where $\omega_x=\omega_y=\omega_z$ produces large degeneracies scaling as $(n+1)(n+2)/2$ for total quantum number $n=n_x+n_y+n_z$ ). Limiting case: letting $k_x=k_y=k_z=k$ recovers $\omega_x=\omega_y=\omega_z=\omega$ and $E_n = \hbar\omega(n+3/2)$ with $n=n_x+n_y+n_z$ , reproducing the standard isotropic-oscillator degeneracy formula $g_n = (n+1)(n+2)/2$ — confirming the general separated-energy machinery correctly reduces to the textbook symmetric result.

Example 3 — Exam/competition level: testing whether a coupled potential separates, and finding the correct coordinates when it doesn't in Cartesian form. A student is given the potential for a particle in 2D,

V(x,y) = \tfrac12 k\left(x^2+y^2\right) + \lambda\, xy

with $k = 400\ \text{N/m}$ and a coupling constant $\lambda = 150\ \text{N/m}$ , and asked whether this separates in $x,y$ , and if not, to find coordinates in which it does.

Step 1 — check Cartesian separability. Written as is, $V(x,y) = \frac12 kx^2 + \frac12 ky^2 + \lambda xy$ has an explicit cross term $\lambda xy$ that cannot be absorbed into a function of $x$ alone plus a function of $y$ alone (no matter how you split it, some residual $xy$ term remains unless $\lambda=0$ ). So the potential is not separable in Cartesian $(x,y)$ .

Step 2 — rotate coordinates. Define rotated coordinates

u = \frac{x+y}{\sqrt2}, \qquad v = \frac{x-y}{\sqrt2}

(a 45° rotation, which leaves the kinetic energy Laplacian form-invariant since it's an orthogonal transformation: $\partial_x^2+\partial_y^2 = \partial_u^2+\partial_v^2$ ). Then

x^2+y^2 = u^2+v^2, \qquad xy = \frac{u^2-v^2}{2}

V(u,v) = \tfrac12 k(u^2+v^2) + \frac{\lambda}{2}(u^2-v^2) = \tfrac12\left(k+\lambda\right)u^2 + \tfrac12\left(k-\lambda\right)v^2

This is additively separable in $(u,v)$ ! Physically, the rotation found the normal coordinates of the coupled oscillator — exactly the same logic used to diagonalize coupled vibrational potentials in normal-mode analysis.

Step 3 — read off the two effective force constants and frequencies.

k_u = k + \lambda = 400+150 = 550\ \text{N/m}, \qquad k_v = k - \lambda = 400 - 150 = 250\ \text{N/m}

With mass $m = 1.00\times10^{-26}\ \text{kg}$ :

\omega_u = \sqrt{\frac{550}{1.00\times10^{-26}}} = \sqrt{5.50\times10^{28}} = 2.345\times10^{14}\ \text{s}^{-1}

\omega_v = \sqrt{\frac{250}{1.00\times10^{-26}}} = \sqrt{2.50\times10^{28}} = 1.581\times10^{14}\ \text{s}^{-1}

Step 4 — total energy in the new separated basis.

E_{n_u n_v} = \hbar\omega_u\left(n_u+\tfrac12\right) + \hbar\omega_v\left(n_v+\tfrac12\right)

For the state $(n_u,n_v)=(2,1)$ :

E_{21} = (1.0546\times10^{-34})\left[2.345\times10^{14}(2.5) + 1.581\times10^{14}(1.5)\right]

= (1.0546\times10^{-34})\left[5.863\times10^{14} + 2.372\times10^{14}\right] = (1.0546\times10^{-34})(8.235\times10^{14})

= 8.684\times10^{-20}\ \text{J} = 0.5421\ \text{eV}

Limiting case: as $\lambda \to 0$ , $k_u, k_v \to k$ and $\omega_u,\omega_v \to \omega=\sqrt{k/m}$ , recovering the ordinary isotropic 2D oscillator with $E = \hbar\omega(n_u+n_v+1)$ — confirming that the rotated-coordinate solution reduces correctly to the uncoupled case, and demonstrating the general principle that separability may require choosing the right coordinate system rather than being an intrinsic yes/no property of the physical system.

Limiting cases, scope & common traps

Situation	What happens	Why
$V=V_x+V_y+V_z$ exactly	Full separation, $E=E_x+E_y+E_z$ , $\Psi=XYZ$	Additive potential + Laplacian already diagonal in these coordinates
$V$ has a cross term ( $xy$ , $r$ -dependence in Cartesian, etc.)	No separation in that coordinate system	Cross term prevents isolating single-variable brackets (Example 3)
Coupled potential quadratic in coordinates	Often separates after rotating to normal coordinates	Orthogonal transformations preserve the Laplacian's form
Central potential $V(r)$	Separates in spherical coordinates, not Cartesian	Spherical Laplacian splits into radial + angular parts; feeds directly into the [[Particle on a Sphere and Spherical Harmonics	particle on a sphere]] treatment
Equal parameters in each direction (cube, isotropic oscillator)	Extra ("accidental" or symmetry-required) degeneracies appear	Permutations of quantum numbers give identical sums
Unequal parameters (asymmetric box/oscillator)	Degeneracies lift; energies become generically distinct	Separated energy formula loses permutation symmetry
$n=1$ (ground states only, one quantum number nonzero)	Reduces to the corresponding 1D zero-point/ground energy	Other separated pieces sit at their own ground state

The chemist behind it

The technique of separation of variables long predates quantum mechanics, developed across the 18th and 19th centuries by mathematicians studying the wave equation, heat conduction, and celestial mechanics — Daniel Bernoulli, Jean le Rond d'Alembert, Joseph Fourier, and Pierre-Simon Laplace all contributed pieces of the method for classical PDEs (the vibrating string, heat diffusion, gravitational potential theory). But it was Erwin Schrödinger who, in his 1926 papers founding wave mechanics, immediately recognized that his new equation for the hydrogen atom was of exactly the same separable type as the classical wave equations he already knew intimately from continuum mechanics and optics.

Origins · 1 minErwin Schrödinger and how this was discovered

Erwin Schrödinger

1887–1961 · Vienna, then Zurich and Berlin, 1920s–1930s

Schrödinger came to quantum theory not as a young revolutionary but as an established expert in classical wave theory, statistical mechanics, and general relativity, already in his late thirties. Working largely in isolation over a few months in early 1926 (famously partly at a mountain retreat), he formulated wave mechanics by demanding that particle mechanics emerge as a short-wavelength limit of a more fundamental wave equation — mirroring how geometric optics emerges from wave optics. His four papers of that year, "Quantisierung als Eigenwertproblem" (Quantization as an Eigenvalue Problem), solved the hydrogen atom essentially by transforming to spherical coordinates and applying exactly the separation-of-variables technique long used for the classical wave equation and Laplace's equation in potential theory — recognizing that the Coulomb potential's spherical symmetry, not its specific $1/r$ form alone, was what made the angular part solvable in complete generality (yielding the spherical harmonics common to any central potential). Schrödinger shared the 1933 Nobel Prize in Physics with Paul Dirac. His broader legacy in chemistry is that every orbital picture — s, p, d shapes, radial nodes, quantum numbers $n,\ell,m$ — descends directly from applying this separation technique to the hydrogen Hamiltonian.

History

- 1926 — Schrödinger publishes wave mechanics and solves the hydrogen atom by separating variables in spherical coordinates.

- 1926 — Schrödinger and independently Heisenberg/Born/Jordan establish the equivalence of wave and matrix mechanics.

- 1927 — Separation of variables becomes a standard tool for the particle in a box, harmonic oscillator, and rigid rotor in early quantum chemistry texts.

- 1930s — Systematic classification of coordinate systems in which the Helmholtz/Schrödinger equation separates (Cartesian, spherical, cylindrical, elliptical, and others) developed in mathematical physics.

- Present — separation of variables remains the entry point for every exactly solvable quantum chemical model taught at the undergraduate level.

Does separation of variables always give the exact solution, or is it an approximation?

When the Hamiltonian is genuinely separable — meaning the potential is additively decomposable in the coordinates used — separation of variables gives the exact solution, not an approximation. The particle in a box, the harmonic oscillator, and the hydrogen atom are all solved exactly this way. The approximation enters only when a real system's potential is not exactly separable (e.g., anharmonic vibrational coupling between normal modes, or electron-electron repulsion in atoms), in which case chemists use a separable reference Hamiltonian as a starting point and treat the residual coupling with perturbation theory or variational methods.

Why doesn't the hydrogen atom separate in Cartesian coordinates, but does in spherical coordinates?

The Coulomb potential is $V(r) = -e^2/(4\pi\epsilon_0 r)$ where $r=\sqrt{x^2+y^2+z^2}$ mixes all three Cartesian coordinates in a way that cannot be split into $V_x(x)+V_y(y)+V_z(z)$ . In spherical coordinates $(r,\theta,\phi)$ , however, $V$ depends on $r$ alone, and the spherical Laplacian itself splits cleanly into a radial second-derivative operator plus an angular operator (built from $\theta,\phi$ ) divided by $r^2$ . This structure allows the wavefunction to be written as $R(r)Y(\theta,\phi)$ , with the angular part further separating into $\Theta(\theta)\Phi(\phi)$ — producing the spherical harmonics central to atomic orbital shapes.

Is separation of variables the same thing as the orbital approximation in atoms?

They are closely related but not identical. Separation of variables is the exact technique used to solve a genuinely separable Hamiltonian (like the one-electron hydrogen atom) into single-particle equations. The orbital approximation in many-electron atoms is a deliberate approximation — the true Hamiltonian is not separable because of electron-electron repulsion, so chemists replace the exact interaction with an effective, separable mean-field potential (as in Hartree-Fock) precisely so that separation of variables can then be applied to the approximate Hamiltonian, yielding one-electron orbitals.

What happens to separation of variables when a coupling term like λxy is present?

Strict Cartesian separation fails whenever the potential contains a genuine cross term that cannot be split into single-variable pieces. However, as shown for the coupled 2D oscillator, a coordinate transformation (often a rotation to normal coordinates) can sometimes restore separability in a new coordinate system. This is the mathematical basis of normal-mode analysis in molecular vibrations: the coupled Cartesian displacement coordinates are rotated (via diagonalizing the mass-weighted Hessian) into normal coordinates in which the vibrational potential becomes a sum of independent harmonic terms.

Why does separability directly explain degeneracy patterns in the particle in a box?

Because the total energy is a sum of independent contributions, $E = E_{n_x}+E_{n_y}+E_{n_z}$ , degeneracy is purely a question of how many distinct quantum-number triples give the same numerical sum. When box edges are equal, permuting quantum numbers leaves the sum $n_x^2+n_y^2+n_z^2$ unchanged, producing symmetry-required degeneracy; sometimes even non-permutation triples coincidentally give the same sum (accidental degeneracy). Making the edges unequal removes the permutation symmetry from the formula, and the corresponding degeneracies are lifted — a direct, checkable consequence of the separated-energy formula.

Does separation of variables apply only to time-independent problems?

The article's focus is the time-independent Schrödinger equation, where separating spatial coordinates gives stationary-state energies and wavefunctions. But the same logical structure also separates time from space in the full time-dependent Schrödinger equation: writing $\Psi(x,t) = \psi(x)e^{-iEt/\hbar}$ is itself an application of separation of variables, with $E$ playing the role of the separation constant linking the time-only equation to the space-only (time-independent) equation. So the technique operates at two nested levels: space-time separation, and then, within the spatial equation, coordinate-by-coordinate separation.

How does this concept connect to the particle on a ring and particle on a sphere?

Both of those models are central-potential-like problems in curvilinear coordinates: the particle on a ring separates naturally using the single angular coordinate $\phi$ , giving quantized angular momentum directly as a separation constant. The particle on a sphere extends this by separating the polar angle $\theta$ and azimuthal angle $\phi$ , producing spherical harmonics as the angular eigenfunctions. Both are direct applications of the same separation-of-variables logic developed here, just carried out in coordinate systems adapted to the geometry (circular or spherical) rather than Cartesian axes.