Aromaticity as a computed shell closure

Six pi electrons in a ring: aromatic. Four: antiaromatic. Ten: aromatic again. The rule is memorised in every first course and derived in almost none, which leaves a very good result looking like a piece of numerology.

It is not numerology, and the derivation is short enough that the omission is hard to defend.

The level pattern is the whole of it #

A monocyclic system’s Hückel eigenvalues have a closed form:

$$x_k = 2\cos\!\left(\frac{2\pi k}{n}\right), \qquad k = 0, 1, \ldots, n-1.$$

Cosine takes the same value at $k$ and at $n-k$. So every level appears twice except the one at $k=0$, and — when $n$ is even — the one at $k = n/2$.

That is the entire derivation. A ring’s orbitals come as one non-degenerate lowest level, then a series of degenerate pairs, and for an even ring one non-degenerate highest level at the top.

Filling that ladder with electrons two at a time, a closed shell needs two electrons for the bottom level and then four for each pair. Two, six, ten, fourteen. Which is $4n+2$.

The rule is therefore a statement about the degeneracy pattern of a cycle, and the degeneracy pattern is a consequence of the ring’s rotational symmetry — a symmetry the exact molecule has too, which is why a rule derived from a matrix of ones and zeroes survives contact with reality.

What the machinery actually does #

The figure above is not produced by evaluating $4n+2$. It is produced by filling.

For each ring size, the eigenvalues are computed, grouped into shells wherever two are equal to within a tolerance, and filled from the bottom. A shell offered more electrons than it can hold takes what it can; a shell offered fewer takes one per orbital before pairing, which is Hund’s rule. The highest shell holding any electrons is then examined, and the report is whether it came out full.

The number $4n+2$ appears nowhere in that procedure. It is asserted afterwards, as a check: the closure the filling found must agree with the arithmetic predicate, for every ring from three to ten, and the build stops if it does not.

That distinction between producing a result and checking one is the whole difference between this figure and a table.

Why 4n is worse than nothing #

The rule’s other half is the interesting one, and it is a genuine prediction rather than a classification.

Two things about that figure deserve attention.

The delocalisation energy is exactly zero, not merely small. Four electrons occupying $x = 2, 0, 0$ give a total pi energy of $4\alpha + 4\beta$, and two isolated ethenes give $4\alpha + 4\beta$. The ring is worth precisely nothing. That is a much stronger statement than “less stabilised than benzene”, and it is arithmetic rather than estimate.

The prediction is a triplet, which is falsifiable and which is wrong. Square cyclobutadiene would have two unpaired electrons. Real cyclobutadiene is a singlet, because it distorts to a rectangle — the two Kekulé-like structures are no longer equivalent, the degeneracy is lifted, and both electrons drop into the lower of the two.

That distortion is a Jahn–Teller effect, and it is exactly the kind of thing a theory that discarded the geometry cannot see. What makes the failure valuable is that it is specific. A theory predicting “cyclobutadiene is unstable” could not be caught out. One predicting a triplet can be, and the mechanism of its being caught out is itself a result.

The odd rings, and the ions that fix them #

The rule as usually stated is about neutral molecules, and neutral molecules are a poor test of it, because most rings’ natural electron counts are wrong.

The ions are the good test, because they let the count be varied while the ring is held fixed.

The cyclopentadienyl anion, with six pi electrons, closes its shell and comes out at 2.47β. The cyclopentadienyl cation, with four, does not. The same carbon skeleton, aromatic one way and antiaromatic the other, which is as clean a demonstration as the subject offers that aromaticity is about a count rather than about a substance.

Eight electrons, and the escape #

Cyclooctatetraene is the case that shows what a molecule does when the count comes out badly and it has room to manoeuvre.

The molecule declines the prediction. Cyclooctatetraene is not planar; it folds into a tub, the p orbitals are no longer parallel, and the pi system becomes four essentially isolated double bonds. Its measured bond lengths alternate — 1.33 and 1.46 ångström — which is what a set of localised double bonds looks like.

That escape is invisible to the theory in the strictest sense. Planarity was assumed before the matrix was written down, so a treatment that produced a tub would have had to be given the geometry it discarded. What Hückel can say is that the planar form is unattractive, and the tub is then the molecule’s answer to a question the theory posed correctly and could not follow up.

The pattern generalises. A molecule facing a 4n count has three ways out — distort within the plane, as cyclobutadiene does; leave the plane, as cyclooctatetraene does; or gain or lose electrons, as the cyclopentadienyl and tropylium systems do. All three are observed, and the count is what makes each of them worth doing.

What was computed, and how #

Every number above comes from diagonalising an adjacency matrix with a cyclic Jacobi sweep, and every one is checked three ways.

The closed form $2\cos(2\pi k/n)$ gives the eigenvalues of any ring exactly, so the solver is compared against it for every size from three to twelve; the worst disagreement is a few parts in $10^{16}$. The two trace relations — the eigenvalues sum to zero, their squares sum to twice the bond count — hold independently of any closed form. And the pairing theorem is asserted in both directions, so an odd ring whose levels came out symmetric about α would be refused.

The delocalisation energies each carry a stated reference. A delocalisation energy is a difference, and the thing it is measured from is recorded in the system’s definition as a count of isolated double bonds — three ethenes for benzene, two for cyclobutadiene, one for cyclopropenyl. Change the reference and the figure prints a different number, which is the honest form of a quantity whose textbook value ranges from 120 to 180 kilojoules per mole depending on the comparison chosen.

The assertion that rejects is the one on shell closure: for every ring from three to ten, the closure found by filling must match the $4n+2$ predicate, and a mismatch stops the build.

The surprise: aromaticity is a property of a graph #

The rule was derived above from nothing but the cyclic connectivity. No bond lengths, no atoms, no energies.

That has a consequence that is easy to miss and hard to unsee. Anything with the right connectivity and the right electron count gets the same answer. Rings of nitrogen, rings of boron and nitrogen alternating, rings of nothing in particular — the level pattern is the cycle’s, and the closure condition is the same.

It also means aromaticity as Hückel defines it says nothing about stability in the ordinary chemical sense. A closed shell means the electron count fits the level pattern. It does not mean the molecule is unreactive, or isolable, or planar, or that it exists.

The gap between the graph criterion and the chemical property is where most of the confusion about aromaticity lives, and naming it as a gap is more useful than trying to close it.

What it costs #

The computation costs nothing: eight small eigenvalue problems, microseconds each.

What the rule costs is stated in what it declines to answer, and the list is longer than a first course suggests.

It cannot say whether a ring is planar, and planarity is a precondition for the p orbitals to be parallel and therefore for the whole treatment to apply. Cyclooctatetraene has eight pi electrons, is antiaromatic by the count, and resolves the problem by folding into a tub — at which point the pi system is a set of isolated double bonds and the question no longer arises.

It cannot rank two aromatic systems against each other in any physically meaningful way, because the energies are in units of an unfitted β.

And it cannot handle a ring with a heteroatom without a fitted adjustment to that site’s α, which is a patch. Pyridine and pyrrole are both aromatic and the reasons differ, and neither reason is in the graph.

Where the model stops #

Three limits, and the third is the one that matters for how the word is used.

Only monocyclic, strictly. The derivation above is about a single cycle. Naphthalene is aromatic and has ten pi electrons, which looks like a confirmation and is a coincidence: its level pattern is not one-and-then-pairs, and applying $4n+2$ to a fused system works often enough to be misleading. The general criterion for a polycyclic system is a computation rather than a count.

Only pi, and only planar. The sigma–pi separation is exact only when a mirror plane puts the two in different symmetry species. A twisted ring has no such plane and the whole framework becomes an approximation of unstated quality.

“Aromatic” means at least four different things. A closed pi shell by this count; an unusual thermodynamic stability; a diamagnetic ring current in a magnetic field; and a tendency to substitute rather than add. The four criteria mostly agree and do not always, and this essay establishes only the first. A source that slides between them without saying so is the most common way the subject is taught badly — the same promotion of a description into a mechanism that the hybridisation account suffers from, met in a different corner of the syllabus.

And the count says nothing about a barrier. Nothing on this site computes an energy surface, so no statement here bears on how fast anything reacts. A closed shell is a statement about a ground state, and chemistry mostly happens on the way to somewhere else.

Who found it, and when #

Hückel derived the rule in 1931, in the same series of papers that introduced the method. It went largely unnoticed for two decades — the name “Hückel’s rule” dates from the 1950s, and Hückel himself was reportedly surprised to find it attached to him.

The delay had a cause worth recording. Hückel’s papers were written for physicists and the rule was buried in them; the chemists who would have used it were reading Pauling. Doering coined the term “aromatic sextet” and popularised the count in the early 1950s, at which point it spread very fast.

The synthesis of the tropylium cation by Doering and Knox in 1954 is what made it stick. A seven-membered carbon ring bearing a positive charge should have been an unpromising thing; it turned out to be remarkably stable, exactly as the count predicted, and the prediction had come from a decades-old paper nobody had read.

The other criteria, and why they mostly agree #

Since the word carries at least four meanings, it is worth asking why they usually coincide — because the coincidence is not obvious and is the reason the concept is useful at all.

The ring current is the criterion a spectroscopist reaches for. A closed shell of delocalised pi electrons circulating round a ring in a magnetic field generates a field of its own, which shifts the resonance of any nucleus in it. Benzene’s protons appear far downfield of an ordinary alkene’s; the protons inside a large annulene appear far upfield, because the induced field points the other way there.

That criterion and the count agree because both depend on the electrons being able to circulate freely round the ring, which requires an unbroken cycle of parallel p orbitals and a shell that is closed. Two different consequences of one structural fact.

The thermodynamic criterion — an unusual stabilisation relative to some reference — agrees for the same reason and is the weakest of the four, because it inherits the reference-state problem entirely. A stabilisation quoted without its reference is not a number.

The reactivity criterion — substitution rather than addition — agrees because addition would break the cycle, and breaking a closed shell costs what the closure gained.

So the four are consequences of one structural fact rather than four independent tests, which explains both why they agree and why the disagreements, when they come, are informative. A system that shows a ring current and no unusual stability is telling something specific about which part of the structural fact holds.

Where the ladder goes next #

The theory that produces the level pattern is Hückel theory, with its checks.

The quantity its eigenvectors give for nothing is bond order, which distinguishes benzene’s uniformity from naphthalene’s.

The general claim about what “delocalised” buys is delocalisation, and the correction to the naive version is delocalisation is not always stabilising.

The reason a ring’s levels pair up at all is a node count meeting a boundary condition.

What the pictures here cannot show. Every level diagram on this page is a ladder of energies in units of β, and β has no number. The vertical spacing is therefore in arbitrary units and no two of these figures can be compared for absolute energy — only the pattern, the degeneracies and the closure are meaningful. Nothing here can show that benzene is chemically unreactive or that cyclobutadiene is difficult to isolate, because those are statements about reaction barriers and no barrier is computed anywhere on this site.