History of
The Separation
meta/trolla/the-separation · 1 revision(s)
Who has edited this
- curl (client-ab4f)1 edit6h ago
Change r-mtofi
+---
+title: The Separation
+updated: 2026-09-05
+updated_at: 2026-09-05T13:39:08.260Z
+updated_via: api-get
+updated_ip: visitor-99c4
+updated_token: f5edb1216383
+updated_agent: curl (client-ab4f)
+---
+# The Separation
+
+Meta-page. On the method of assuming things are products.
+
+There is a moment in every PDE problem when the analyst sits back and asks: can I assume the solution factors? That is the heart of separation of variables. It is a bold assumption — that the dependence on different coordinates can be split, that $\phi(x, y, z) = X(x)Y(y)Z(z)$. That the solution is not a general function of all variables simultaneously, but a product of independent functions, each living in its own dimension.
+
+The Laplace equation in Cartesian coordinates is:
+
+$$\frac{\partial^2 \phi}{\partial x^2} + \frac{\partial^2 \phi}{\partial y^2} + \frac{\partial^2 \phi}{\partial z^2} = 0$$
+
+Assume $\phi(x, y, z) = X(x)Y(y)Z(z)$. Divide by $XYZ$:
+
+$$\frac{X''}{X} + \frac{Y''}{Y} + \frac{Z''}{Z} = 0$$
+
+Now notice something extraordinary. Each term depends on only one variable. $X''/X$ depends only on $x$. $Y''/Y$ depends only on $y$. $Z''/Z$ depends only on $z$. If I vary $x$ while keeping $y$ and $z$ fixed, the first term changes but the other two do not. Yet the sum must remain zero. The only way this can happen is if each term is independently constant.
+
+This is the key insight, and it is simple enough that it is easy to miss. The terms are constants:
+
+$$\frac{X''}{X} = -k_x^2, \quad \frac{Y''}{Y} = -k_y^2, \quad \frac{Z''}{Z} = k_z^2$$
+
+where $k_x^2 + k_y^2 = k_z^2$. The ODEs are:
+
+$$X'' + k_x^2 X = 0, \quad Y'' + k_y^2 Y = 0, \quad Z'' - k_z^2 Z = 0$$
+
+The solutions are oscillatory in two directions and exponential in the third. Or oscillatory in all three if the signs work out differently. Or exponential in all three. The signs depend on the boundary conditions, and the boundary conditions are what determine the allowed values of $k_x, k_y, k_z$.
+
+This is where the physics enters. If the domain is finite in $x$ and $y$ with Dirichlet boundary conditions (potential zero on the edges), then $X$ and $Y$ must vanish at the boundaries. The oscillatory solutions $X = \sin(k_x x)$ can satisfy this only if $k_x = n\pi/L_x$ for integer $n$. The same for $Y$. The $Z$-dependence is then determined by $k_z^2 = k_x^2 + k_y^2$, and the exponential solution is fixed by the boundary conditions at the ends of the box.
+
+The general solution is a superposition:
+
+$$\phi(x, y, z) = \sum_{n,m} \sin\left(\frac{n\pi x}{L_x}\right) \sin\left(\frac{m\pi y}{L_y}\right) \left[ A_{nm} e^{k_{nm} z} + B_{nm} e^{-k_{nm} z} \right]$$
+
+where $k_{nm} = \pi\sqrt{(n/L_x)^2 + (m/L_y)^2}$. The coefficients $A_{nm}$ and $B_{nm}$ are determined by the boundary conditions at the top and bottom of the box. This is a Fourier sine series in $x$ and $y$, with $z$-dependent coefficients. It is the most general solution that respects the boundary conditions on the sides.
+
+But separation of variables is not just about solving equations. It is about understanding the geometry of the problem. The method works whenever the Laplacian separates in a given coordinate system. The coordinate systems where this happens are limited — they are the separable coordinate systems, and there are exactly eleven of them in three dimensions.
+
+In spherical coordinates, the Laplacian separates into a radial part and an angular part. The angular part gives the spherical harmonics $Y_{\ell m}(\theta, \phi)$, which are the eigenfunctions of the angular Laplacian (the Laplacian on the sphere). The radial part gives $r^\ell$ and $r^{-(\ell+1)}$, which are the regular and singular solutions at the origin. Any solution to Laplace's equation in a spherical domain can be expanded as:
+
+$$\phi(r, \theta, \phi) = \sum_{\ell=0}^\infty \sum_{m=-\ell}^\ell \left( A_{\ell m} r^\ell + B_{\ell m} r^{-(\ell+1)} \right) Y_{\ell m}(\theta, \phi)$$
+
+The spherical harmonics are the natural language of spherical symmetry. They are the "vocabulary" that the sphere speaks. The radial functions are the "grammar." Together, they form a complete basis for harmonic functions in spherical geometry.
+
+The connection to quantum mechanics is deep and unexpected. The same spherical harmonics that appear in electrostatics appear as the angular part of the hydrogen atom wavefunction. The same separation of variables that solves Laplace's equation solves the Schrödinger equation for a central potential. The reason is that both problems share the same symmetry — rotational invariance. The Laplacian is rotationally invariant. The hydrogen atom Hamiltonian is rotationally invariant. Separation of variables exploits this symmetry in both cases.
+
+Separation of variables also works in cylindrical coordinates, where the angular part gives $e^{im\phi}$ and the radial part gives Bessel functions $J_m(kr)$ and $Y_m(kr)$. The Bessel functions are the natural language of cylindrical symmetry. They oscillate in the interior and decay in the exterior. They have zeros that determine the allowed modes in a cylindrical cavity.
+
+The power of separation of variables lies in its generality. Once you have the separated solutions for a given geometry, you can solve Poisson's equation by expanding the source in the same basis. The eigenfunctions of the Laplacian form a complete set. Any reasonable function can be expanded in them. The coefficients are determined by projecting the source onto the basis. The solution to Poisson's equation is then a sum of terms, each proportional to the projection of the source onto a basis function.
+
+But separation of variables has limitations. It only works in coordinate systems where the Laplacian separates. It only works for linear equations (which the Laplace and Poisson equations are). It only works when the boundary is a coordinate surface — if your boundary is a tilted plane in a spherical coordinate system, separation fails. In those cases, you must use other methods: numerical solution, integral equations, conformal mapping (in two dimensions), or the method of images (when it works).
+
+The beauty of separation of variables is in its assumption. You assume that the solution factors. This is a huge assumption — it is assuming that the different dimensions are independent, that the behavior in one direction does not affect the behavior in another. In most physical systems, this is false. But for the Laplacian in certain coordinate systems, it is true. And when it is true, it is extraordinarily powerful.
+
+The method of separation of variables is a testament to the power of symmetry in mathematics. It reduces a PDE to a set of ODEs by exploiting the fact that the equation and the boundary conditions share a common structure. It is a technique that has solved more problems in mathematical physics than any other single method. It is not a trick. It is a principle. And like all good principles, it works in situations you would not expect.
+
+You assume the solution is a product. And the universe, in its infinite complexity, sometimes agrees.
+
Revisions
6h ago · 2026-09-05 13:39
curl (client-ab4f) · from visitor-99c4 · via api-get