Where the atoms go

The carbonyl pins nothing the closure counts

The rotamer ceiling on the gem-dimethyl effect rests on a count of frozen rotations, derived from the open chain as n − 3. The obvious worry about a real lactonisation was the carbonyl: a torsion that prefers to be planar might pin one rotor and lower every ceiling. Built with the hydroxy acid's own bonds and angles, the chain that closes runs from the hydroxyl oxygen to the carbonyl carbon, and the carbonyl's torsion turns the chain's last bond, which moves no atom of it. The count stays n − 3 and the verdicts do not move by a digit. What the carbonyl adds is a direction — an attack near 107° to the C=O bond — and that same free rotation supplies it from every closable five- and six-ring conformer, at every bond angle tried. A torsion pinned inside the chain does take a rotor away, exactly one.

Worth reading first: The ring closes from the arrangement that alternates · The count that was never written down.

Two methyl groups on one carbon of a chain make the chain close into a ring faster, by factors up to eleven thousand for γ-butyrolactone, and the explanation usually given has the wrong sign for every ring from five up. The rotamer account replaces it with a count: substituents that make anti torsions expensive confine the chain to its gauche states, from which closure is easier, and the largest acceleration that can buy is a ceiling that rises with ring size, ((1+2x)/2x)r((1 + 2x)/2x)^r with x=e−g/kTx = e^{-g/kT} and rr the number of rotations the closure freezes.

The count rr was a convention for most of this argument — three of them, never written down — until the ring closes from the arrangement that alternates derived it. Build the open chain’s staggered conformers in three dimensions, find the ones whose ends come close enough to close, and read rr off the ceiling those conformers give: it is n−3n - 3 exactly, because every conformer closer than the first one with an anti torsion is all gauche.

That chain had equal bonds of 1.53 Å and equal angles of 112°. A real lactonisation does not. The chain that closes is a hydroxy acid, with an oxygen at one end and a carbonyl at the other, and the essay ended by naming the worry about that carbonyl: its torsion prefers to be planar rather than staggered, a torsion pinned to planar values would take a rotor out of the count, and that would lower every ceiling by a factor of 3.3.

It does not, and the reason is about which bonds belong to the chain.

Which atoms the closing chain contains

A γ-hydroxy acid closes to γ-butyrolactone when its hydroxyl oxygen attacks its own carboxyl carbon. The five atoms that become the ring are the hydroxyl oxygen, the three carbons between, and the carbonyl carbon. The chain whose ends must meet is those five atoms, and its two ends are the oxygen that attacks and the carbon attacked.

The torsions that move those ends relative to each other are the rotations about the chain’s internal bonds — for five atoms, the two bonds between the second and fourth. The rotation about the bond from the last methylene to the carbonyl carbon turns the carbonyl group, its oxygen and its hydroxyl, about the line of that bond. The carbonyl carbon lies on the axis of that rotation, so turning it moves no atom of the chain. It moves the carbonyl’s own oxygen, which is not one of the chain’s atoms and not one whose distance decides closure.

So the planar preference of the carbonyl torsion, whatever its size, applies to a torsion the count never included. The count of n−3n - 3 for a chain of nn atoms counts the rotations about bonds two to n−2n - 2, and the carbonyl’s bond is bond n−1n - 1.

The chain with its own bonds and angles

The argument from which bonds are internal settles the pinning question without numbers. The ester’s other differences from an equal-bond chain still need checking, because they move the conformers’ end-to-end distances, and the count depends on which conformers lie below the first one with an anti torsion.

Built with the hydroxy acid's own bonds and angles, every chain still closes only from all-gauche conformers below its first anti one. The hydroxyl oxygen to carbonyl carbon distance of every staggered conformer of the chains that close five-, six- and seven-membered lactones, with conformers having an anti torsion marked apart; the line marks each chain's first conformer with an anti torsion, and the tick the same for a chain of equal bonds and angles. 5 atoms: closest +− at 2.63 Å (equal bonds 2.69), first anti 4.32 Å (equal bonds 4.44); 6 atoms: closest +−+ at 1.80 Å (equal bonds 1.79), first anti 3.97 Å (equal bonds 4.05); 7 atoms: closest +−+− at 0.41 Å (equal bonds 0.32), first anti 2.74 Å (equal bonds 2.78).
Fig. 1 The distance from the hydroxyl oxygen to the carbonyl carbon in every staggered conformer of the chains that close five-, six- and seven-membered lactones, built with the ester’s bond lengths and angles. The dashed line marks each chain’s first conformer with an anti torsion; the tick, the same for the equal-bond chain.

The chain is built with an O–C bond of 1.43 Å, C–C bonds of 1.53 Å and a C–C(=O) bond of 1.51 Å, an O–C–C angle of 109.5° and C–C–C angles of 112°, and every internal torsion staggered at 180° or ±60°. The distances move little. For five atoms the closest conformer is the alternating one at 2.63 Å, against 2.69 on the equal-bond chain, and the first conformer with an anti torsion moves from 4.44 to 4.32 Å. For six, 1.80 against 1.79 for the closest and 3.97 against 4.05 for the first anti. For seven, 0.41 against 0.32 and 2.74 against 2.78. The shorter O–C bond and the smaller angle at the oxygen bring the oxygen in slightly, and the window below the first anti conformer narrows by about a tenth of an ångström on each chain.

The ordering that mattered survives. On every chain the closest conformer is the one whose gauche torsions alternate in sign, and every conformer below the first anti one is all gauche. So the derived count is n−3n - 3 on the ester chain exactly, checked at the bottom, middle and top of each chain’s window for rings of five to eight, and the ceilings are the ones the equal-bond chain gave: 10.99 for the five-membered ring and 36.46 for the six, at butane’s gauche energy and 298 K.

The refusal: a torsion that is pinned

“The carbonyl pins nothing” would be an empty statement if the derivation could not have seen a pinned torsion where there was one. So the same chain is built with one internal torsion confined to the two planar values, 0° and 180°, instead of the three staggered ones — what a carbonyl or an ester linkage in the interior of the chain would do, as in a closure onto an oxygen beyond the carbonyl.

A torsion pinned inside the chain removes one rotor; the carbonyl's own torsion is not inside the chain. The effective rotor count the ceiling implies for the 6-membered closure, across the closure distances below the chain's first anti conformer (1.82 to 3.95 Å): for the ester chain by distance, by distance and attack direction, and for the same chain with one internal torsion confined to 0° and 180°. ester chain: 3.000; ester, attack direction too: 3.000; one internal torsion planar: 2.000.
Fig. 2 The effective rotor count the ceiling implies for the six-membered closure across its window: the ester chain by distance alone, by distance and attack direction, and the same chain with one internal torsion confined to 0° and 180°.

The pinned chain’s count is n−4n - 4 exactly: 1.000 for five atoms, 2.000 for six and 3.000 for seven, across the window and whichever internal torsion is pinned. A planar torsion contributes no gauche weight either way, so the substituent that penalises anti torsions has nothing to act on there, and the ceiling loses precisely the factor that torsion would have contributed. The derivation sees a pinned rotor when one exists, to the last digit. On the lactone’s own chain there is none to see.

What the carbonyl does add: a direction

The carbonyl is not irrelevant to closure; it is irrelevant to the count. A nucleophile does not attack a carbonyl carbon from any direction. Crystal structures in which an amine nitrogen or an oxygen sits near a carbonyl carbon put the approach at about 107° to the C=O bond, out of the carbonyl’s plane — the Bürgi–Dunitz angle, read from the geometry of molecules caught at every stage of approach. An oxygen close to the carbonyl carbon but on the wrong side, along the C=O bond or directly behind the carbon, is not in a position to react.

The direction of the C=O bond is exactly what the carbonyl’s free rotation sets. The C=O bond makes 56° with the extension of the last C–C bond (the C–C=O angle is 124°), so as the carbonyl turns, the C=O direction sweeps a cone of half-angle 56° around that line. If the line from the carbonyl carbon to the hydroxyl oxygen makes an angle β\beta with the same axis, the angle between it and the C=O bond runs over

θ∈[ ∣56∘−β∣,  56∘+β ]\theta \in \big[\,|56^\circ - \beta|,\; 56^\circ + \beta\,\big]

as the carbonyl turns, and a conformer can present the Bürgi–Dunitz direction if 107° lies in that range.

The 5-membered closure's nearest conformer, with the carbonyl turned to meet the hydroxyl at 107°. The +− conformer of the chain from the hydroxyl oxygen to the carbonyl carbon for a 5-membered lactone, built with the ester's bond lengths and angles and projected on its two widest axes. The hydroxyl oxygen is 2.63 Å from the carbonyl carbon; the line to it makes 95° with the extension of the last C–C bond, so turning the carbonyl sweeps the attack angle from 39° to 151°, and it is drawn at 107°.
Fig. 3 The five-membered closure’s nearest conformer, projected on its two widest axes, with the carbonyl turned so that the line to the hydroxyl oxygen meets the C=O bond at 107°.

For the five-membered ring’s closest conformer the oxygen sits at β=95°\beta = 95°, and the carbonyl’s rotation sweeps the attack angle from 39° to 151°. It can be turned to 107°. The picture above is drawn with it turned there: the hydroxyl oxygen 2.63 Å from the carbonyl carbon, off to one side of the C=O bond, out of the carbonyl’s plane.

Turning the carbonyl reaches 107° from every closable five- and six-ring conformer; the seven-ring's nearest is at the edge, and the eight-ring's comes from behind. For each all-gauche conformer closer than its chain's first anti one (one of each mirror pair), the range of angles between the C=O bond and the line to the hydroxyl oxygen that the carbonyl's rotation sweeps, against 107°: 5, +−: 39–151°; 5, ++: 70–180°; 6, +−+: 19–131°; 6, +−−: 64–176°; 6, ++−: 27–139°; 7, +−+−: 5–107°; 7, ++−+: 31–143°; 7, +−−+: 31–143°; 7, +−++: 82–180°; 7, ++−−: 36–148°; 8, +−+−+: 117–180°; 8, ++−+−: 5–117°.
Fig. 4 For each all-gauche conformer closer than its chain’s first anti one, the range of attack angles the carbonyl’s rotation sweeps, against 107°.

Every closable conformer of the five- and six-membered chains has a range containing 107°: 39° to 151° and 70° to 180° on the five-ring, 19° to 131°, 64° to 176° and 27° to 139° on the six. The seven-ring’s other closable conformers reach it comfortably, but its nearest, the fully alternating one, reaches it only at the very edge of its range, 5° to 107°: its oxygen lies at β=51°\beta = 51°, almost along the line of the last C–C bond, and the carbonyl can be turned to meet it at 107° and no wider. The idealised eight-atom chain’s nearest conformer is the one case where the direction fails outright: its oxygen sits behind the carbonyl carbon at β=173°\beta = 173°, and the attack angle can only range from 117° to 180°.

A ceiling that does not care which conformers close

The direction adds a condition to closure, and it would be natural to expect it to change the ceiling wherever it removes a closable conformer. It does not, and the reason is in the ceiling’s form.

The ceiling is the closable fraction of the all-gauche chain divided by the closable fraction of the plain chain. When every closable conformer is all gauche, the closable weight is the same in the numerator and the denominator and cancels, and what is left is the plain chain’s whole Boltzmann weight divided by its all-gauche weight — ((1+2x)/2x)n−3((1 + 2x)/2x)^{n-3}, whichever conformers close. The ceiling counts rotations frozen, not conformers closed. A condition that removes some closable conformers and leaves at least one changes nothing, so long as everything that remains is all gauche.

That is why the seven-ring’s marginal conformer does not matter to its ceiling, and why on the five- and six-membered rings the question did not even arise: the direction removes nothing there. The eight-atom chain is the exception that shows the rule’s boundary. At closure distances below 1.47 Å its only closable conformer is the one that approaches from behind, the direction removes it, and the chain has no closable conformer at all — a ceiling with nothing to divide by. That chain’s closest approach of 1.15 Å is an artefact of perfect staggering in a ring that large, and real eight-membered closures go through conformers this enumeration does not hold, but it marks the one way a direction could ever enter the count: by removing every way to close.

Robust to the angles that were chosen

The ester’s angles are not known to a degree. The O–C–C angle at the hydroxyl-bearing carbon, the C–C–C angles along the chain, and the C–C=O angle at the carbonyl all vary by a few degrees between structures and with substitution, and a gem-dimethyl group opens its own C–C–C angle.

Across twenty-seven sets of angles, the five- and six-ring conformers always reach 107°, and the seven-ring's often does not. For each of 27 combinations of the O–C–C angle (107°, 109.5°, 112°), the C–C=O angle (120°, 124°, 128°) and the C–C–C angle (110°, 112°, 114°), the smallest upper end of the attack-angle range among each chain's closable conformers. 5 atoms: 144° to 157°, every conformer reaching 107° in 27 of 27; 6 atoms: 122° to 139°, every conformer reaching 107° in 27 of 27; 7 atoms: 98° to 127°, every conformer reaching 107° in 16 of 27.
Fig. 5 Across 27 combinations of the O–C–C, C–C–C and C–C=O angles, the largest attack angle the worst-placed closable conformer of each chain can reach, against 107°.

Across every combination of an O–C–C angle of 107°, 109.5° or 112°, a C–C=O angle of 120°, 124° or 128°, and C–C–C angles of 110°, 112° or 114°, the five-membered chain’s worst-placed closable conformer still reaches at least 144°, and the six-membered chain’s at least 122°: every closable conformer of both presents the Bürgi–Dunitz direction under every one of the 27. The seven-membered chain’s nearest conformer reaches 107° under 16 of them and falls short, by as much as 9°, under the other eleven. The window below the first anti conformer moves between 4.22 and 4.41 Å for five atoms and between 3.79 and 4.13 for six, and the count is n−3n - 3 inside it every time.

The reason the small rings are so forgiving is visible in the formula for the range. Its width is set by β\beta: when the hydroxyl oxygen lies roughly broadside to the last C–C bond, β\beta near 90°, the range runs from about 35° to about 145° and contains any plausible attack angle. A five- or six-membered chain holds its closable oxygens near that broadside arrangement — β\beta between 75° and 126° in every one of them — because a chain that short cannot reach far round the carbonyl in either direction. It is only when a chain is long enough to fold its oxygen round behind or in front of the carbonyl — β\beta near zero or near 180° — that the cone can miss, and in the staggered enumeration that first happens at seven atoms.

The verdicts on the measured accelerations do not move: the ester chain gives the same ceilings to the last digit. Each measured gem-dimethyl acceleration of lactone formation against the rotamer ceiling at a closure distance of 3.2 Å: on the equal-bond chain, on the ester chain, and on the ester chain with the attack direction required. γ-butyrolactone formation, gem-dimethyl, 250: 10.99, 10.99, 10.99; γ-butyrolactone formation, tetramethyl, 11000: 10.99, 10.99, 10.99; δ-valerolactone formation, gem-dimethyl, 10: 36.46, 36.46, 36.46.
Fig. 6 Each measured gem-dimethyl acceleration of lactone formation against the rotamer ceiling on the equal-bond chain, on the ester chain, and on the ester chain with the attack direction required.

So the verdicts do not move, by any digit. The 250-fold gem-dimethyl acceleration of γ-butyrolactone formation is 22.7 times the five-ring’s ceiling of 10.99, and the 11,000-fold tetramethyl acceleration a thousand times it: the rotamer account is refuted on the five-membered ring with the ester’s own geometry exactly as with the idealised one. The 10-fold acceleration of δ-valerolactone formation is below the six-ring’s 36.46: not refuted, and still not supported either, since the gauche energy’s own quoted range moves that ceiling past the measurement.

How the chains were built

Each chain is built atom by atom from its bond lengths, its angles and its torsions, by the natural extension of a reference frame, exactly as the equal-bond chain was; with every bond 1.53 Å and every angle 112° the construction reproduces that chain’s end-to-end distances to 10−910^{-9} Å. Every internal torsion is 180° or ±60°, so a chain of nn atoms has 3n−33^{n-3} conformers. A conformer’s Boltzmann weight is e−gk/RTe^{-g k/RT} for kk gauche torsions at butane’s gauche energy of 3.8 kJ/mol and 298 K — the two enter only as their ratio, so a different temperature is a different gauche energy and changes nothing below that is not already a statement about xx; the substituted chain is the limit in which every anti torsion is forbidden. The ceiling is computed as the closable fraction of the all-gauche chain over that of the plain chain, and the effective count as its logarithm in units of ln⁡((1+2x)/2x)\ln((1 + 2x)/2x).

The attack range is the closed form above: the angle between the C=O bond and the line to the hydroxyl oxygen, as the C=O bond turns on its cone about the last C–C bond. The picture of the five-ring conformer places the carbonyl oxygen 1.21 Å from the carbon at the rotation that brings that angle nearest 107°, by a scan in half-degree steps. The pinned chain replaces the staggered values of one internal torsion by 0° and 180° and gives that torsion no gauche weight.

What a staggered hydroxy acid cannot say

A carbonyl rotation is not free in energy. The rotation about a carboxylic acid’s Cα–C bond has a preferred orientation and a small barrier, and turning it to present the attack direction costs something. That cost is not in the count, and it would matter if it differed between the substituted and unsubstituted chains — which it might, since a gem-dimethyl group next to the carboxyl changes what the carbonyl eclipses. On the γ-hydroxy acid the substituents are two carbons away from it.

Staggered torsions are an idealisation. Every conformer here has torsions of exactly 180° or ±60°. A real chain relaxes each torsion off those values, and a closable conformer is a region of torsion space rather than a point. The derivation’s claim is that the count is a property of which kind of conformer closes, and that survives any relaxation that does not move a conformer with an anti torsion below the all-gauche ones.

The direction is a crystal-structure average. The Bürgi–Dunitz angle is a statistic of molecules caught near a carbonyl, not a requirement with a sharp edge. A tolerance of ten degrees either side changes none of the five- or six-membered results, and changes which of the seven-membered chain’s geometries count as reaching it.

Nothing here is a rate. The ceiling bounds what freezing rotations can buy. The measured accelerations include whatever the gem-dimethyl group does to the transition state, and a refutation of the rotamer account says only that freezing rotations cannot be the whole of it.

The count was about the chain between the two atoms

What the derivation counts is the number of rotations that move one end of the closing chain relative to the other. The carbonyl’s planarity, the ester’s shorter bonds and its smaller angle at the oxygen are all real differences between a hydroxy acid and an idealised chain, and none of them changes that number, because none of them changes which bonds lie between the two atoms that must meet. The direction of attack is a real requirement too, and it is met by the one rotation that is free precisely because it lies outside the chain.

The atoms that meet across a ring showed two closed conformers differing more in a contact than in an energy, and this is the same lesson from the other side: the geometry that decides a ring’s thermodynamics and the geometry that decides its closure are different parts of the molecule.

Still open: a contact with a sign, and a carbonyl that is not free

The obvious open question is still the contact. The alternating conformer carries the syn-pentane contact between atoms four apart, priced so far as a repulsion. In a hydroxy acid the closest such contact in the closable conformers is the hydroxyl oxygen’s own approach to the carbonyl carbon, which is the first stage of a bond, not a clash. An attraction of 2.9 kJ/mol a pair would sink the six-membered verdict, and what that contact is worth needs an energy of the two groups as what they are, which nothing here supplies.

The nearer question is the carbonyl’s rotation, now that it is known to carry the direction. Its preferred orientation in the open acid, and the barrier to turning it, would put a weight on each closable conformer’s attack — and if the gem-dimethyl group changes that preference, it would act on the closure through the one rotation the count has just been shown not to contain.

Shares its objects with

Essays naming at least two of the same things, that neither author linked.

Named objects

A dashed tag is an object no other essay names yet.

ClosureConformerRing strainRotamerTorsion