An angular ring is priced by its neighbours
Worth reading first: The angular factor is a ratio of two prices · The angular factor belongs to the chain.
A magnetic flux through one ring of a fused chain of benzene rings changes the current in every other, and how much depends on what lies between the two rings. The scatter counts angular rings found that each angular ring between a pair cuts the response by a factor; the angular factor belongs to the chain found that factor falling with the chain’s length; and the essay that gave each passage its own price showed the old factor to be a ratio of two prices — the share of the response kept across each linear ring and across each angular ring — and both of them moving. The linear price rises from 0.465 at seven rings to 0.551 at thirteen, as the acenes’ gap closes. The angular price falls from 0.305 to 0.275.
Except in the zigzag. Fitted on the zigzag chain alone, where every ring but the two ends is angular, the angular price rises with length, 0.318 to 0.337. Two passages of the same kind, priced on different chains, move in opposite directions. That essay named the obvious difference between them: how much linear chain surrounds each angular ring. In the zigzag an angular ring’s neighbours are angular rings; in a chain with one or two angular rings among linear ones, its neighbours are linear. It asked whether an angular passage is cheaper or dearer according to its company.
It is, by a factor of more than two, and the two trends are the two ends of one dependence.
Three angular passages
Every angular ring strictly between a pair has two neighbours along the chain, and each is either a linear ring or another angular ring. So there are three kinds of angular passage: with no linear neighbour, inside a run of angular rings; with one, at the edge of a run; and with two, an isolated angular ring among linear ones. The model keeps everything else the passage model had — one constant, one log price per linear passage, one log factor per angular ring the pair stands on — and replaces the single angular price with three.
The chains are the same design as before: at each of seven, nine, eleven and thirteen rings, every chain with at most two angular rings wherever they are, every run of three adjacent angular rings, the straight chain and the zigzag — 77 chains and 4,389 pairs at thirteen. The responses are the same second cross-derivatives of the Hückel π energy with respect to fluxes through two rings, the same pairs and the same values; nothing new is solved. At thirteen rings the angular passages between pairs divide into 230 with no linear neighbour, 472 with one and 1,135 with two.
The three prices are far apart. At thirteen rings an angular ring with no linear neighbour keeps 0.443 of the response, one with one keeps 0.278 and one with two keeps 0.192. An angular ring among angular rings costs barely more than a linear ring, 0.443 against 0.571; an angular ring between two linear ones costs three times as much. The single angular price, 0.275, sits almost exactly on the middle class, because the pairs that fixed it were a mixture of all three and mostly of the third.
The three also move differently with length. The no-neighbour price rises, 0.377, 0.407, 0.429 and 0.443; the one-neighbour price falls slightly, 0.311 to 0.278; the two-neighbour price falls steeply, 0.251 to 0.192. The linear price rises throughout, a little faster than the no-neighbour angular price, from 0.473 to 0.571. Measured against it, the no-neighbour price is a nearly fixed fraction, 0.80 of the linear price at seven rings and 0.78 at thirteen, as though a run of angular rings were a slightly lossier version of a linear stretch with the same length dependence. The two-neighbour price is not: it falls from 0.53 of the linear price to 0.34, and the old angular-ring factor, itself a ratio of an angular price to the linear one, was dominated by the isolated angular rings that make up most of its passages.
The puzzle was one dependence
The zigzag’s rising price is the no-neighbour price, diluted. A zigzag of thirteen rings has eleven angular rings, of which nine have only angular neighbours and the two at the ends of the run each have one linear neighbour, the terminal ring; its own price, 0.337 at thirteen rings, sits between the no-neighbour 0.443 and the one-neighbour 0.278, and rises with the first. The single price fitted across every chain falls because most of its angular passages are isolated angular rings, the two-neighbour kind, whose price falls. Nothing about the zigzag was special except its company.
That makes the length dependence itself a statement about company. A ring current carried along a run of linear rings has the acenes’ character: a small gap that closes with length, and a response that decays more slowly the longer the run. A run of angular rings has the phenacenes’ character, with a large gap and a short decay. What the numbers say is that an angular ring’s passage price is not a property of the ring; it is a property of the stretch of chain the response is travelling through, and an angular ring embedded in linear chain inherits some of the linear chain’s length dependence in the opposite direction — it becomes a sharper break the more the chain around it lengthens.
Is the gain the split, or just more parameters?
Three prices instead of one is two more parameters, and two more parameters always fit better. The question is whether they fit better because they describe which rings are which, or merely because there are more of them.
The passage model explains 93.7 to 93.8 per cent of the variation in the log responses at every length. The three-price model explains 94.7, 95.5, 96.0 and 96.3 per cent: at thirteen rings it removes two-fifths of what was left. The control deals each pair’s three counts to a different pair with the same total of angular rings, so the model has exactly the same parameters and the same number of angular passages per pair, and only which passages are which is scrambled. Averaged over three shuffles it explains 93.8, 94.0, 94.3 and 94.6 per cent: a twelfth of the real gain at seven rings, rising to a third at thirteen. The improvement belongs to the neighbours.
The control’s own small gain is not nothing, and it has a plain source: the shuffle keeps each pair’s total, and pairs with many angular passages are mostly in chains with runs, so even scrambled counts carry a little of the chain type. The real split carries all of it.
One factor per linear neighbour
Plotted on a logarithmic scale against the number of linear neighbours, the three prices at each length lie close to a straight line. So the three-price model can be replaced by a two-parameter one: an angular price for a ring with no linear neighbour, multiplied by a fixed factor for each linear neighbour it has. That model’s is within 0.0001 of the three-price model’s at every length. The factor per linear neighbour is 0.815 at seven rings, 0.753 at nine, 0.705 at eleven and 0.666 at thirteen.
That factor carries the whole length dependence of the angular price’s company. An angular ring between two linear rings at thirteen rings is priced as an angular ring among angular rings times 0.666 squared, and the square is why the two-neighbour price falls so much faster than the one-neighbour price. The factor itself is falling at about the rate the linear price is rising: from seven to thirteen rings the linear price rises by a factor of 1.21 and the factor per linear neighbour falls by 1.22. Whether that is a coincidence of four lengths or a single number seen twice is the obvious thing to ask next.
Chain by chain
The model is fitted on all the pairs at once, and an average over chains could hide chains that disagree with it. Each mixed chain at thirteen rings can be fitted alone with one angular price, and that price set against the mean company of its own angular rings.
The chains fall into three groups and the groups are ordered. Runs of three angular rings, whose mean company is two-thirds of a linear neighbour, have prices of 0.278 to 0.302; pairs of adjacent angular rings, with one linear neighbour each, 0.250 to 0.286; and chains whose angular rings are isolated, with two linear neighbours each, 0.157 to 0.227. Within each group the price varies with where the angular rings sit, symmetrically about the chain’s middle: highest when they are next to an end — 0.302 for a run of three beside the terminal ring, 0.223 for an isolated angular ring there — and lowest in the middle, 0.278 and 0.157. That is the end effect the earlier essays found in another form, and it is the part of the variation one ring’s company cannot reach. Between groups the ordering is the company’s, and no chain is out of place.
What the split takes from the other factors
A model’s coefficients share out the variation between them, and splitting one moves the others. Two moved here, and both in a direction that says something.
The factor for each angular ring a pair stands on, rather than passes through, falls when the angular passage is split: from 0.681 to 0.640 at seven rings and from 0.548 to 0.471 at thirteen. Under the single price, part of what an end ring’s company does was being absorbed by the factor for standing on an angular ring, because the pairs that stand on angular rings in runs are the pairs whose passages are cheap. With the company priced separately, standing on an angular ring is dearer than it looked, and its own length dependence is steeper: it falls by a factor of 0.74 from seven rings to thirteen, against 0.80 before. The pair’s own rings and the rings between it are not separable as cleanly as the passage model wrote them, and the split has moved part of the burden from one to the other.
The linear price moves too, up by about four per cent at every length — 0.473 against 0.465 at seven rings, 0.571 against 0.551 at thirteen. Under the single angular price a linear ring next to an angular ring was charged a little of its neighbour’s cost; with the angular ring’s company priced, that cost goes where it belongs, and the linear passage looks slightly cheaper. The separation that neither distance nor fusion count could explain alone is now carried by three kinds of passage and one kind of ending. What the model still lacks is visible in the chain-by-chain prices: the same company priced higher next to a chain’s end than in its middle, by about a tenth for a run of three and by two-fifths for an isolated angular ring.
How the prices were fitted
Every response is the one the passage-price essay used: for each pair of rings, the second derivative of the Hückel π energy of the fused chain with respect to magnetic fluxes through the two rings, at half filling with a uniform hopping, on chains built from their lists of angular rings with the angular rings read back from the geometry. Each angular ring between a pair is classed by its two neighbours along the chain, the rings on either side of it in the sequence of fusions, and a neighbour beyond the chain’s end does not exist and does not count. The log responses are fitted by ordinary least squares on the named counts, with the response kept per passage the exponential of each coefficient.
The checks run where the figures are drawn. The passage model refitted here reproduces the published linear and angular prices at every length to . On every pair the three counts add to its number of angular rings. The three-price model contains the passage model, so its is never lower. And the refusal: the same counts shuffled among pairs that share their total of angular rings, under three different shuffles, must buy less than half of the real gain at every length. They buy at most a third.
What a Hückel chain cannot say
One electron and one hopping. Every response here is a Hückel response with equal hopping on every bond, so bond alternation, which is what makes an angular ring’s outer bonds short and its inner ones long in a real phenacene, is absent. Bond alternation would change the angular ring’s passage in exactly the direction its company does, by localising its π electrons, and the separation into company and alternation is not something this model can make.
Neighbours, not neighbourhoods. The class counts only the two rings adjacent to an angular ring. The chain-by-chain residuals within each group show that rings further along the chain matter too, and a model with a reach longer than one ring is one parameter away.
Least squares on logarithms. Every price is fitted to the log response with equal weight on every pair, so pairs far apart, whose responses are small and whose logarithms are large and negative, count as much as near pairs. A fit weighted towards the pairs a measurement could actually see would move the prices, probably towards the near pairs’ values, and the company’s effect is largest there.
A design with few long runs. The no-neighbour class at thirteen rings rests on 230 passages, most of them in the zigzag and in runs of three; its price is the least constrained of the three. A design with runs of four and five angular rings would test whether it stays a single number or changes along a longer run.
A passage is a property of the stretch
The sequence that measured these responses began by asking whether a current divides at a fusion, and then found that a reach has no length when the gap is everything. Each step since has found a quantity that looked local — a factor per angular ring, a price per passage — and then found it depending on something further away: the chain’s length, then the kind of passage, now the passage’s neighbours. The common thread is that a ring current is a property of the π system as a whole, and an angular ring’s effect on it is the effect of a change in the chain’s character at that point. Among angular rings there is no change and the price is mild; between linear rings the change is abrupt and the price is steep.
Which says something about the habit of reading aromaticity ring by ring. Turning one fusion changes a whole chain’s responses through one integer, the number of angular rings; here the same turned fusions set how strongly a ring passes on a current induced elsewhere, and the setting is not by the ring that is turned but by the pair of rings beside it.
Still open: whether the two length dependences are one, and a longer run
The obvious open question is the coincidence above. The linear price rises by 1.21 from seven to thirteen rings and the factor per linear neighbour falls by 1.22. If the factor is the linear price’s reciprocal times a constant — if a linear neighbour costs an angular ring exactly what the linear passage beside it gains — then the whole model has one length-dependent number, the gap of the linear stretches, and every price follows from it. A fit with the factor tied to the linear price, against the free fit, at fifteen and seventeen rings where the lever is longer, would settle it.
The nearer question is the run. Runs of four and five angular rings, at nine to thirteen rings, would say whether an angular ring deep inside a run is cheaper still than one with angular neighbours on both sides, or whether the no-neighbour class is already the limit — whether company reaches past the nearest ring.
What links here
Computed from the collection rather than written here: the essays that point at this one.
Reads more easily once this is understood
Essays that name this one as worth reading first.
Shares its objects with
Essays naming at least two of the same things, that neither author linked.
- The floor was in the bookkeeping — both name band gap, hückel theory, least-squares
- The reach is the molecule's — both name band gap, hückel theory, least-squares
- A decay that keeps slowing down — both name band gap, least-squares
- A parameter that never finds a value — both name hückel theory, least-squares
- A stabilisation is measured from somewhere — both name aromaticity, hückel theory
- Aromaticity as a computed shell closure — both name aromaticity, hückel theory
Named objects
A dashed tag is an object no other essay names yet.