Beyond the octet

Half a per cent of an axial bond

A steep contact between five ligands prefers the square pyramid by an amount that sets the whole case for one, and every calculation of it put all five ligands at one distance. Give the bipyramid's two axial bonds their own length and the advantage turns out to be worth at most half a per cent of that length, at any power of the contact. Phosphorus pentafluoride's axial bonds are 2.8 per cent longer than its equatorial ones and phosphorus pentachloride's 5.1. At measured lengths no power law makes a square pyramid, unless the size is held by a different convention, and then the verdict flips.

Worth reading first: Priced against the whole bond · Only a steep contact makes five square.

A σ-only bond energy has the trigonal bipyramid as the only minimum for five ligands at a main-group centre, and pentaphenylantimony is a square pyramid anyway. The repair that worked was a contact between the ligands steeper than the fifteenth power of their separation, because five points repelling each other change their own best arrangement from the bipyramid to the square pyramid at an exponent of 15.05. The price of that contact turned out to be large, a seventh of the σ bond energy under the thirtieth power and more than the whole bond under the sixteenth. The reason it was large was that near its own threshold a steep law tells the two shapes apart by only a few per cent of itself.

Every one of those calculations put the five ligands on one sphere. The σ energy in this model is a function of the five directions alone, so for the bonding the assumption costs nothing. The contact is the one term that acts on the ligands as atoms, at their positions, and a real bipyramid does not have one bond length. In phosphorus pentafluoride the axial bonds are 1.577 Å and the equatorial 1.534, by gas electron diffraction. In phosphorus pentachloride they are 2.124 and 2.020. And the molecule that started this, pentaphenylantimony, has an apical bond of 2.115 Å and four basal bonds averaging 2.216 in its square-pyramidal crystal.

So the open question was what a second length does to an argument whose margin was already a few per cent. The answer is that it does everything.

What the angular advantage is worth in bond length

The cleanest way to put the question is to ask how much longer the bipyramid’s two axial bonds must be, with the equatorial bonds fixed, before the contact alone can no longer tell it from the best square pyramid. Call that elongation k∗−1k^*-1. Below the Riesz exponent there is nothing to cancel, because the contact already prefers the bipyramid, and the elongation is zero. Above it, the square pyramid has an angular advantage, and some lengthening of the bipyramid’s longest-reaching contacts must eventually pay for it.

The whole angular advantage of the square pyramid is worth half a per cent of an axial bond. Against the exponent s of a 1/rˢ contact between ligand atoms, the elongation of the bipyramid's two axial bonds at which the contact can no longer tell it from the best square pyramid. With equatorial and basal bonds at one length the elongation rises from zero at the Riesz exponent, 15.05, to 0.53 per cent near s = 46 and falls back. With the mean of the five bonds held instead it peaks at 2.60 per cent. Measured bipyramids are elongated by 2.8 (PF₅) and 5.1 per cent (PCl₅).
Fig. 1 The axial elongation at which a power-law contact cannot tell the bipyramid from the best square pyramid, against the exponent, under two conventions for size.

With the four basal bonds of the pyramid and the three equatorial bonds of the bipyramid held at one length, the cancelling elongation rises from zero at s=15.05s=15.05 to 0.053 per cent at the sixteenth power, 0.23 at the twentieth, 0.46 at the thirtieth, and peaks at 0.53 per cent near s=46s=46. Then it falls again, to 0.42 at the hundredth power and 0.16 at the four-hundredth. At no exponent does it reach six tenths of a per cent.

That fall at large exponents is the steep limit doing what a steep limit does. As ss grows a power law sees less and less of the arrangement and more and more of its shortest distances. The bipyramid’s shortest contacts are its six axial–equatorial pairs at 2\sqrt2 times the bond. Five points on a sphere cannot all be further apart than that, and the square pyramid at its best does not manage it either. An elongated bipyramid’s six pairs, though, are longer than 2\sqrt2, and nothing in the pyramid is. So under a steep enough law any elongation at all wins, and the curve has to come back towards zero.

Measured bipyramids are not near this curve. PF₅’s axial bonds are 2.8 per cent longer than its equatorial ones, more than five times the peak; PCl₅’s are 5.1 per cent longer, nearly ten times. So with the equatorial length held, a power-law contact at measured lengths prefers the bipyramid at every exponent.

Why the number is so small

The margin between the shapes and its sensitivity to length can be read separately, and they are of very different sizes.

Under a steep contact every tenth of a per cent of axial length moves the verdict by more than the angles do. The logarithm of the best square pyramid's contact energy over the bipyramid's, against the elongation of the bipyramid's axial bonds, with equatorial and basal bonds at one length, for four exponents. Below zero the pyramid is preferred. At equal lengths every curve starts below zero by its angular advantage, −0.023, −0.069, −0.132, −0.209, and each rises with a slope near half its exponent, because the six axial–equatorial distances grow by half the elongation and carry most of the bipyramid's contact energy. The steeper the law, the larger its advantage and the faster it is spent: they cross zero at 0.23, 0.46, 0.53, 0.42 per cent.
Fig. 2 The logarithm of the best square pyramid’s contact energy over the bipyramid’s against the axial elongation, at four exponents, equatorial and basal bonds at one length.

At equal lengths the logarithm of the ratio of the two contact energies is the square pyramid’s angular advantage. It is −0.023-0.023 at the twentieth power, −0.069-0.069 at the thirtieth, −0.13-0.13 at the fiftieth and −0.21-0.21 at the hundredth: the steeper the law, the more clearly it prefers the pyramid. Each line then rises with the elongation, and the slopes are 9.9, 15.0, 25.0 and 50.0. The slope is half the exponent, to within two per cent at every one.

The reason is geometry with no model in it. An axial–equatorial distance with the axial bond lengthened by a factor kk is 1+k2\sqrt{1+k^2} times the equatorial bond, and its logarithmic derivative at k=1k=1 is exactly one half. Under 1/rs1/r^s an energy dominated by those six pairs therefore falls as s2\tfrac s2 times the elongation, and the cancelling elongation is the advantage divided by that:

k∗−1≈2 δ(s)s,k^*-1\approx\frac{2\,\delta(s)}{s},

where δ(s)\delta(s) is the angular advantage. Every number above follows from it: 2×0.023/202\times0.023/20 is 0.23 per cent, 2×0.132/502\times0.132/50 is 0.53. And it says why there is a peak. The advantage grows with the exponent, but more slowly than the exponent does past s≈46s\approx46, so dividing one by the other gives a curve that rises and then falls.

So the steeper laws, the ones that prefer the square pyramid most firmly, are also the ones most sensitive to bond length, in exactly the proportion that cancels their firmness. A contact steep enough to choose the square pyramid clearly is a contact for which a fifth of a per cent of axial bond is as large as the whole difference between the shapes.

The six contacts that decide

It helps to see which distances are moving.

The six contacts that decide a steep contact are exactly the six an axial elongation lengthens. The ten ligand–ligand distances, in units of the equatorial bond, for the bipyramid at equal bond lengths, for the bipyramid with PCl₅'s measured axial ratio, and for the best square pyramid under a twentieth-power contact. The bipyramid's six closest contacts are its axial–equatorial ones, at √2. PCl₅'s ratio lengthens all six by 2.6 per cent, which under the twentieth power multiplies their energy by 0.60. The square pyramid's four closest, at 1.404, are untouched by it.
Fig. 3 The ten ligand–ligand distances, in units of the equatorial bond, for the bipyramid with equal bonds, the bipyramid with PCl₅’s axial ratio, and the best square pyramid under the twentieth power.

The bipyramid at one length has its six axial–equatorial contacts at 2\sqrt2, three equatorial contacts at 3\sqrt3 and one axial pair at 2. The best square pyramid under the twentieth power has an apex angle of 96.9°, four basal contacts at 1.404, shorter than anything in the bipyramid, four apex–basal contacts at 1.497 and two diagonals near 1.99. Its advantage is that it has traded six contacts at 2\sqrt2 for four slightly closer and four noticeably further.

Give the bipyramid PCl₅’s axial ratio and its six closest contacts move out to 1.451, a 2.6 per cent increase, which under the twentieth power multiplies their energy by 0.60. Nothing about the pyramid’s geometry has changed, and the pyramid’s whole angular advantage at the twentieth power was a factor of 0.977. A molecule’s own measured bond lengths move the decisive contacts by twenty times what the angles were worth.

That is also where the essay on five-coordinate site sizes found its result. A repulsion between sites on a bipyramid, given one longer bond, puts it axial, because the axial site is the crowded one, and the long bond goes to the crowded site for that reason. The same crowding that makes the axial site the place for a long bond makes the axial bonds’ length the variable a steep contact is most sensitive to. A contact steep enough to make the square pyramid is precisely the kind of repulsion that, given any freedom in the lengths, would stretch the bipyramid’s axial bonds and so remove the square pyramid it was introduced to make.

With the σ energy back

The contact alone decides only which shape a contact would choose. The case for pentaphenylantimony needed the contact to overcome the σ energy too, and that was a share: the bipyramid’s contact energy, at the weight where the square pyramid takes over, over its σ cost of one sixth.

With the σ energy back, a tenth of a per cent of axial length ends the sixteenth power's square pyramid. At q = 2, a main-group centre, the contact share at which the square pyramid takes over — the bipyramid's contact energy over its σ cost of one sixth — against the axial elongation, for four exponents. Under 1/r¹⁶ it runs from 1123 at equal lengths to 20218 at 0.05 per cent, and no share does it from 0.10. Under 1/r²⁰ it runs from 233 at equal lengths to 1652 at 0.20 per cent, and no share does it from 0.25. Under 1/r³⁰ it runs from 101 at equal lengths to 785 at 0.40 per cent, and no share does it from 0.50. Under 1/r⁵⁰ it runs from 69 at equal lengths to 1274 at 0.50 per cent.
Fig. 4 At q = 2, the contact share at which the square pyramid takes over against the axial elongation, for four exponents.

At q = 2, a main-group centre, the share at equal lengths is 1,123 under the sixteenth power, 233 under the twentieth, 101 under the thirtieth and 69 under the fiftieth, the numbers that set the price. Each rises as the axial bonds lengthen, and each stops existing. Under the sixteenth power a twentieth of a per cent takes the share from 1,123 to 20,220, and a tenth of a per cent ends it. Under the twentieth the share is 1,652 at 0.2 per cent and nothing switches at 0.25. Under the thirtieth it lasts to 0.4 per cent. Under the fiftieth it reaches 1,274 at 0.5 per cent and, by the contact-alone curve, ends just past it. The numbers at q = 3 run about thirty per cent higher and end at the same elongations, since where a share stops existing is a property of the contact alone.

So the share can no longer be read as “large but finite”. It was already a seventh of the bond or more, and that was the value at a length ratio no molecule has.

The exponential contact

A power law is a stand-in, and the earlier work also tried a contact that decays exponentially between ligand atoms 2.2 Å from the centre, the Sb–C length. There the switch needed a range below 0.215 Å, much shorter than a Born–Mayer range near 0.3, which was already a difficulty.

An exponential contact makes the pyramid over a band of ranges that shrinks from both ends as the axial bonds lengthen. For an exponential contact between ligand atoms 2.2 Å from the centre, the ranges at which the contact alone prefers the square pyramid, at six axial elongations with equal equatorial and basal bonds. At equal lengths the band runs from zero range to 0.215 Å; at half a per cent it has narrowed to 0.045–0.102 Å. A Born–Mayer range, about 0.3 Å, is outside it at every elongation.
Fig. 5 For an exponential contact between ligand atoms 2.2 Å out, the band of ranges in which the contact alone prefers the square pyramid, at six axial elongations.

At equal lengths the band runs from zero range to 0.215 Å, because a short-ranged exponential is just a very steep law. Lengthen the axial bonds and the band closes from both ends. The upper edge comes down, as the power laws’ did, and a lower edge appears, because at very short range the exponential sees only the shortest contacts, and an elongated bipyramid’s shortest contacts are longer than the pyramid’s. At 0.1 per cent the band is 0.005 to 0.193 Å, at 0.3 per cent 0.018 to 0.151, at 0.5 per cent 0.045 to 0.102, and at 0.55 per cent there is no band at all. The exponential closes where the power laws’ peak was, at the same half a per cent, because over the distances that matter a short exponential is a power law with an exponent near 2 L/ρ\sqrt2\,L/\rho, and every exponent has been tried.

Which length is held

Everything so far held the equatorial and basal bonds at one length and let the axial ones grow. That is one convention for “the same size”. It is not the only one, and the model has nothing in it that chooses. The σ energy here depends only on directions, and nothing prices a bond length against anything else.

The natural alternative is to hold the mean of the five bond lengths, so an elongated bipyramid has slightly shorter equatorial bonds to pay for its long axial ones. On the curve in the first figure that convention is the upper line: it peaks at 2.60 per cent, because shortening the equatorial bonds brings the six decisive contacts back in. PF₅’s 2.8 per cent is still just past it and PCl₅’s 5.1 well past.

The real test is to give both shapes their measured ratios, and the only square pyramid with a measured ratio in this argument is pentaphenylantimony, with an apex 4.6 per cent shorter than its base.

With measured bond lengths, which length is held fixed decides whether any contact makes a square pyramid. For three bipyramids (equal lengths, PF₅'s axial ratio, PCl₅'s) against two square pyramids (equal lengths, SbPh₅'s shorter apex), the exponents at which a power-law contact alone prefers the pyramid, under two conventions for the same size. With equatorial and basal bonds held at one length no measured combination gives the pyramid at any exponent. With the mean of the five held, SbPh₅'s apex gives it from s = 20.3 against PCl₅'s ratio and from 8.4 against equal lengths.
Fig. 6 The exponents at which a power-law contact alone prefers the square pyramid, for three bipyramids against two square pyramids, under two conventions for size.

With the bases held, the shortened apex only makes the pyramid worse, since its four apex–basal contacts move in. No combination of measured ratios gives the pyramid at any exponent. With the mean held, the shortened apex lets the four basal bonds grow, and the four basal contacts, the pyramid’s closest, move out. Against equal lengths it gives the pyramid from s=8.4s=8.4, below the Riesz exponent, and against PF₅’s and PCl₅’s ratios from 13.2 and 20.3.

So at measured lengths the verdict of a steep contact is a statement about a convention. Holding the bases says the square pyramid needs something other than a contact. Holding the mean says that, with pentaphenylantimony’s own apex, almost any steep contact gives it. Neither is wrong, and the gap between them is the size of the effect the model was built to find. A model with one length never had to choose, and that is why it could quote a threshold.

What it means for pentaphenylantimony

None of this says why pentaphenylantimony is a square pyramid. It is a square pyramid in one crystal and close to a bipyramid in a cyclohexane solvate, which already said that the two shapes are close in energy and that something outside the isolated molecule decides. What this does say is that the steep-contact explanation, as stated, assumed what it needed. Its margin over the bipyramid is smaller than the length differences every real five-coordinate molecule carries. The prediction of a threshold at the fifteenth power, and the price at the twentieth, belong to molecules with all five bonds equal, and there are none.

The obvious defence is that the lengths are themselves the result of the same forces, so they cannot be taken as given. That is right, and it makes the problem harder, not easier. A model that let the lengths respond would need a σ energy that knows about lengths, a stiffness for each bond and a contact, and on what the crowded-site argument already found, the contact alone would stretch the axial bonds of a bipyramid. Whether the σ energy’s own preference for shorter bonds restrains that by enough to matter is exactly the question the length-free model could not ask.

The deeper surprise is in the slope. Shape arguments are usually made with angles and corrected with lengths, on the reasoning that angles are what a point group cares about. Under a steep contact the order inverts: the angles are a correction of a few hundredths in a logarithm, and a one per cent change of length is a change of s/200s/200 in the same logarithm. A steep law is nearly a hard-sphere law, and for hard spheres shape is decided by who touches whom, which is a question about lengths.

How the comparison was made

The σ energy is the fourth-order form of the five-ligand trade between balance and isotropy, a function of the five directions and the parameter qq, unchanged. The contact is summed over the ten ligand pairs at their actual positions, the directions scaled by each ligand’s own bond length. The bipyramid’s directions are its ideal ones, axial along zz and equatorial at 120°, and the square pyramid’s apex angle is optimised by ternary search at every comparison, under whatever contact and lengths apply. A cancelling elongation is a bisection on kk for a sign change in the log ratio; a window of exponents is a scan from 5 to 600 in steps of one per cent with each sign change bisected; a share is a bisection on the weight for a change of the lower shape, as before. At k=1k=1 the contact alone reproduces the Riesz exponent of 15.05, and at zero weight the σ margin is the σ-only branch difference to 10−910^{-9}, so the lengths are the only thing added. The refusal is a shortened axial bond, which has to help the pyramid, and does: at k=0.995k=0.995 its window opens below 15.05.

Where the model stops

The lengths are inputs. Nothing here computes a bond length. The measured ratios are used as given, and the curves are drawn against a ratio rather than a molecule. A model with relaxed lengths would need a σ energy that depends on them, which this one does not have.

Two conventions, and no reason to prefer either. Holding the bases and holding the mean are the two simplest ways to compare different shapes at “the same size”. A third, such as holding the σ bond energy, would need the missing length dependence again.

Point ligands. A phenyl ring is a flat object, and its contact with its neighbours depends on how the rings turn, which a contact between points cannot represent. That limit was already stated for the equal-length model, and this result makes it more important: if what decides is who touches whom, the shape of what touches is the next thing to know.

σ only. The π rows a ligand carries are still absent.

Who measured what

The pentahalide bond lengths are gas-phase electron diffraction values, PF₅’s from the work of Hansen and Bartell and PCl₅’s from Adams and Bartell; pentaphenylantimony’s square pyramid is the X-ray structure of Beauchamp, Bennett and Cotton. That a trigonal bipyramid’s axial bonds are longer than its equatorial ones for main-group centres is standard structural chemistry. Hypervalent bonding without d orbitals gives the usual reason, the axial three-centre bond, and the count of what s and p can reach is how this argument has set it out. The Riesz problem for five points is the mathematics of energies on a sphere. What is computed here is the length scale: the cancelling elongation under every power, its half-the-exponent slope, and the verdict at measured lengths under both conventions.

Still open: lengths that respond, and a contact with a shape

The obvious open question is the relaxed problem. Give each bond a stiffness and let the σ energy depend on length through its couplings, then minimise over lengths and angles together. The contact would then lengthen the axial bonds, as the site-size argument found it does, and the question is whether the σ energy holds the lengthening below the half a per cent that decides, or whether the bipyramid’s own response to the contact is enough to make the square pyramid unreachable. That needs one coupling-against-length law, the one thing this model has never had.

The nearer question is the ligand’s shape. If the decisive quantity is who touches whom, a phenyl ring’s contact depends on its orientation, and a square pyramid’s four basal rings can tilt in ways a bipyramid’s cannot. A contact between rigid rods, or between pairs of points a fixed distance apart along each bond, would let orientation enter at the cost of one more parameter, and would say whether the square pyramid’s advantage lives in its angles at all or in what its ligands are allowed to do with their room.

What links here

Computed from the collection rather than written here: the essays that point at this one.

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.

ApproximationBond lengthCoordination numberHypervalencyModel limitTrigonal bipyramid