Nodes

An orbital is a function, and a node is a surface on which it is exactly zero — and stays zero, rather than passing through zero on the way somewhere. The number of them is fixed by the quantum numbers and by nothing else.

An orbital with quantum numbers $n$ and $l$ has exactly $n - l - 1$ radial nodes and exactly $l$ angular ones. Both counts are exact, both are computable, and together they are the sharpest check available on whether an orbital has been drawn correctly.

Two kinds #

A radial node is a sphere. The wavefunction changes sign as the radius crosses it, so the orbital has an inner region of one sign and an outer region of the other. A 2s orbital has one; a 3s has two.

An angular node is a surface through the nucleus — a plane, or a pair of cones. The wavefunction changes sign as a direction crosses it. A p orbital has one, which is the plane between its lobes; a d orbital has two.

The two kinds arise from the two factors of the wavefunction. The radial part $R(r)$ contributes the spheres and the angular part $Y(\Omega)$ contributes the surfaces through the origin, and because the function is a product, it vanishes wherever either factor does.

Why the counts are what they are #

The radial function for given $n$ and $l$ is an exponential times a polynomial in $r$ of degree $n - l - 1$, and a polynomial of that degree has that many positive roots here. The angular function for given $l$ is a spherical harmonic, whose nodal structure is $l$ surfaces.

Adding them gives $n - 1$ nodes in total for every orbital of principal quantum number $n$, however the total is split between the two kinds. A 3s has two radial and none angular; a 3p has one and one; a 3d has none and two. All three have two.

That is a genuinely useful pattern. It says the total node count is a property of the shell and the split is a property of the shape, and it explains why higher-$l$ orbitals of the same shell are the ones that look simple.

What a node is not #

Three misreadings, and the third is the one that produces real confusion.

Not a place the electron avoids. The density is zero on the node, which is a set of measure zero — a surface has no volume, so the probability of finding the electron exactly there is zero for the same reason it is zero for any exact position.

Not a barrier. Nothing prevents the electron from being on either side. The question “how does it get from one lobe to the other” presupposes a trajectory, and there is none.

Not the same as a small value. A node is exactly zero, and the sign changes across it. A wavefunction that becomes very small without changing sign has no node there, and the distinction is what makes the count exact rather than approximate.

Counting them, and the bug that found #

The counts can be checked by walking the computed function, and doing it properly was harder than it looks.

Radial nodes are easy: step outward and count sign changes, ignoring the far tail where the function has decayed into arithmetic noise and a spurious flip means nothing.

Angular nodes are not. The first version walked one great circle across the sphere and counted sign changes, which works for a p orbital and fails badly for others. A circle chosen in the plane containing both nodal planes of a $d_{xy}$ orbital touches them only tangentially — the function is proportional to $\cos^2 t$ along it and never goes negative — so the count came out zero for an orbital with two.

The fix is to walk many randomly oriented circles and take the largest count, since a generic circle cuts each nodal surface exactly twice. That version gets every orbital right.

It is worth dwelling on what that error looked like. The drawing was unchanged, the shape was correct, and the reported node count was wrong by two. Nothing about the picture would have revealed it, which is exactly the situation independent checks exist for.

What it costs to count rather than recall #

Recalling a node count costs nothing: $n - l - 1$ radial and $l$ angular, two expressions that fit in a line and are always right. Counting them costs a great deal more and is done anyway, and the reason is worth stating plainly rather than as a slogan.

The radial count walks a few hundred points along a ray and looks for sign changes, with the far tail discarded where the function has fallen into arithmetic noise. That threshold is the one judgement call in it, and it has to be a judgement call: an exponentially decaying function crosses zero spuriously as soon as its magnitude reaches the last bits of a double, and a counter that trusted those crossings would report a different answer at every distance it was asked to look.

The angular count is the expensive one — many randomly seeded great circles, each sampled finely enough to resolve two crossings of every nodal surface, with the maximum taken. That costs perhaps ten thousand evaluations per orbital, which is nothing in absolute terms and is thousands of times the cost of writing $l$.

What the expenditure buys is a test the formula cannot provide. $n - l - 1$ is a statement about the quantum numbers; walking the function is a statement about the radial polynomial that was actually implemented. When the two disagree, the polynomial is wrong — and a wrong Laguerre recurrence produces a function with the right symmetry, the right decay and a completely convincing picture. The counter is the only thing that would notice.

That is the argument for testing any formula, met in its smallest instance. The formula is the claim. The count is the test. A check that merely restated the formula would reject nothing, so the count is also required to fail when it is handed an orbital claiming the wrong number.

What the node does to the picture #

A radial node has a consequence for the drawn surface that almost every published picture leaves out.

If the contour level is low enough, the isosurface has a component inside the node and another outside it: a shell inside a shell. That is the correct picture of a 2s orbital at any sensible level, and the single ball usually drawn is what the surface looks like only when the contour is high enough to have lost the inner region entirely.

The site checks this directly: the number of shells in the drawn surface must be one more than the number of radial nodes. A 2s surface claimed to be a single shell throws.

Nodes and energy #

There is a pattern worth knowing that connects nodes to something measurable.

More nodes means higher energy. A function with more sign changes has more curvature, curvature is kinetic energy, and so within a given atom the orbitals sort by total node count. For hydrogen the energy depends on $n$ alone, which is the same statement, since the total node count is $n-1$.

For a many-electron atom the degeneracy breaks and the s orbital of a shell lies below the p, which lies below the d. The usual explanation is penetration — the s orbital has density close to the nucleus where the shielding by inner electrons is incomplete — and the radial distribution is where that becomes visible.

Nodes in a molecule #

The idea transfers directly and it is where it earns its keep.

When two atomic orbitals combine, the antibonding combination has a node between the nuclei that neither parent had. That node is the whole reason the combination is higher in energy: density has been expelled from the region between the nuclei, where it would have been attracted to both.

So counting nodes in a molecular orbital is counting antibonding character, and the ordering of a molecular orbital diagram is very largely a node count. Overlap decides which combinations form; the nodes decide which of them is which.

The levels, counted rather than described #

A conjugated ring is where that claim can be checked at every level at once, because its orbitals can be produced exactly. The Hückel matrix of a pi system is its adjacency matrix — one where two carbons are bonded, zero everywhere else — and its eigenvectors are the coefficients of the molecular orbitals.

A node in such an orbital is a bond across which the coefficients change sign, which makes it something to count rather than something to see.

Drag the slider up through the levels and the two numbers in the readout move together: the energy rises in units of β, and the node count rises with it. Zero nodes at $\alpha+2\beta$; two nodes for each of the degenerate pair at $\alpha+\beta$; four for the pair at $\alpha-\beta$; six — every bond — for the top orbital at $\alpha-2\beta$.

That is the particle-in-a-box pattern arriving in a molecule. More nodes, higher energy, no exceptions; and the ordering of the whole diagram falls out of a count that requires no picture to make.

It also settles a question about degeneracy that the delocalisation essay leans on. Benzene’s orbitals come in pairs because a ring has two ways to place the same number of nodes — one rotated from the other — while a chain has one. The pairing is a consequence of the node count meeting a boundary condition, and it is what makes a closed shell take two electrons and then four at a time.

What the count cannot tell #

Two limits.

It does not fix the positions. The count is exact; where the radial nodes sit depends on $Z$ and on the details of the radial function, and for a many-electron atom on the approximation used.

It is a property of the one-electron function, and the same caution applies to every quantity here. In a many-electron atom there are no exact orbitals, so the node count is a property of whichever approximate orbitals have been chosen. It happens to be robust — every reasonable approximation gives the same counts — but it is not a measured quantity.

The total is the number that decides the energy #

The two counts are given separately — $n - l - 1$ radial and $l$ angular — and their sum is worth writing down, because it is the quantity hydrogen’s energy actually depends on.

$$(n - l - 1) + l = n - 1.$$

The $l$ cancels. Every orbital of a given $n$ has exactly $n-1$ nodes in total, however they are divided between radial and angular, and hydrogen’s energy depends on $n$ alone.

So the degeneracy of the 2s with the 2p, and of the 3s with the 3p and the 3d, is a statement that nodes cost the same whether they are radial or angular. A 2s has one radial node and no angular ones; a 2p has no radial node and one angular one; they have the same total and the same energy.

That is not a general fact about quantum mechanics. It is a peculiarity of the potential.

The reason nodes cost anything at all is kinetic: a wavefunction that changes sign has to curve, curvature is kinetic energy, and more nodes means more of it. That much is general. What is special about a $1/r$ potential is that the trade between the two kinds of node comes out exactly even — a radial node and an angular node cost precisely the same, so only their sum matters, and the energies collapse into shells labelled by one integer.

Change the potential and the accident goes. In a many-electron atom the inner electrons screen the nucleus, the potential is no longer $1/r$, and a radial node stops costing what an angular one costs — which is exactly why the 2s and 2p separate, and why the periodic table has the shape it has rather than being organised in shells of $n$ alone.

That gives the node count its real standing in this collection. It is a count, exact and unchanged by any approximation, and it is the count the energy depends on in the one case where an exact answer exists. Everywhere else the count survives and the degeneracy does not, and the difference between the two is the whole of what screening does.

The direction the splitting takes follows from the same argument, and it is worth having because it is the fact the periodic table is built on. A radial node sits at a definite distance from the nucleus, and an orbital with one has amplitude on both sides of it — including inside, close to the nucleus, where the screening from the inner electrons has not yet taken effect. An angular node is a plane or a cone through the nucleus, and an orbital with one has no amplitude at the nucleus at all.

So of two orbitals with the same total node count, the one with more of its nodes radial reaches further in, feels more of the nuclear charge, and comes out lower. That is why the 2s lies below the 2p, the 3s below the 3p below the 3d, and the ordering within a shell runs by $l$ — one comparison of two kinds of node, with the same arithmetic that made them degenerate in hydrogen deciding the order everywhere else.

It also names the quantity: an orbital’s penetration is a property of how many of its nodes are radial, which is $n-l-1$. Reading that count is reading how far inside the screening the orbital reaches, which is a great deal more than a check on a drawing.

Where the idea came from #

Nodes are older than quantum mechanics by a long way. They belong to the physics of standing waves, and Chladni was scattering sand on vibrating plates to make them visible in the 1780s.

That lineage is not decorative. The hydrogen wavefunctions are the standing waves of a particular three-dimensional problem, the spherical harmonics were developed by Laplace and Legendre for gravitation, and the mathematics arrived a century before the application. When Schrödinger solved the hydrogen atom in 1926 the functions were already known objects; what was new was what they were for.

The node count also carries a historical irony. Spectroscopists had labelled series of lines sharp, principal, diffuse and fundamental decades before anybody knew what distinguished them — and those labels became s, p, d, f, which is to say the letters now encoding the angular node count were chosen to describe the appearance of spectral lines.

The whole set, counted #

A summary table is worth having, because the counts are simple and the pattern is not obvious until it is written out.

Every entry was produced by walking the computed function rather than copied, and the totals are what make the pattern visible: every orbital of principal quantum number n has n − 1 nodes, however they are split.

That total is the same as the number of nodes of a one-dimensional oscillator’s nth state, and it is not a coincidence: both are counting the sign changes of the nth solution of a Sturm–Liouville problem, and the theorem that the nth eigenfunction has n − 1 nodes is general. The hydrogen atom is a standing-wave problem, and its node counting is the standing-wave node counting.

Where to read on #

The consequence for the drawn surface is what the contour encloses, where the level decides whether the inner shell of a 2s appears at all.

The related question the node count does not answer is where the electron actually is.

And the place nodes do physical work is overlap, where a node between two nuclei is what makes a combination antibonding.

What the pictures here cannot show. A node is a surface of exactly zero, and a drawing shows a contour of some non-zero level — so the node appears as a gap between shells rather than as itself. The claim that the function is exactly zero there is arithmetic, not something visible in any figure.