What symmetry decides

The partner reads the other displacement

Ammonia's E block has a second force field that gives NH₃'s bands exactly, and deuterium can only disfavour it twofold. Its symmetry labels' verdicts are not new numbers: under the partner, a least-cost distortion of the bond angle goes into the degenerate bend in exactly the share that a one-coordinate distortion does under the published field, 0.926 against 0.883 — across the nine-tenths line that decides whether the label settles the coordinate. What separates the two fields is the Coriolis constant: ζ₄ is −0.732 under one and −0.518 under the other, a gap larger than the whole band of fields the frequencies allow.

Worth reading first: Two Coriolis constants are one number · Deuterium breaks the pyramid's partner.

A symmetry label makes a claim about a distortion: that it goes into the vibration of the label’s species and not into the others. Three ways to stretch a bond measured that claim as a share, the concentration of a unit distortion in the labelled vibration, and found that the share depends on what “a distortion” means. Three definitions were tried: the B-matrix row, which moves a central atom as far as a ligand; the least-cost displacement, which disturbs the kinetic energy least; and the one-coordinate displacement, which changes one internal coordinate and holds the rest. A label settles a coordinate when the share is above nine tenths.

The worst case was the loosest constant then showed the share depends on the force field as well. Walked along the stretch–bend constant, the direction its frequencies see least, ammonia’s field moves its bond-angle verdict from 0.840 to 0.923 without the fit getting measurably worse. Two Coriolis constants are one number found the measurement that sees mode shapes directly, the Coriolis constant of each degenerate vibration, and showed that the sum rule ties the two into one number: a measured ζ4\zeta_4 would locate the field within the band sixteen times more sharply than the frequencies do.

All of that was inside one family of fields. Deuterium breaks the pyramid’s partner found another kind of freedom altogether. Each of ammonia’s two symmetry blocks has an exact partner field at NH₃’s masses, the rescaled exchange of its stretch and bend combinations, and ND₃ disfavours the E partner only twofold in band-centre misfit. The partner is not on the walked band; it is a second field elsewhere. So the question is what it does to the verdicts the band was measured on, and which measurement can see it.

The partner exchanges two definitions

The E partner trades the two definitions of a distortion: its least-cost verdicts are the published field's pure ones. For the degenerate stretch and bend of NH₃, the share of a unit distortion that goes into the E vibration its label names, under three definitions of the distortion, for the published field and its E partner. stretch, least-cost: 0.9261 against 0.8832; stretch, pure coordinate: 0.8832 against 0.9261; bend, least-cost: 0.8832 against 0.9261; bend, pure coordinate: 0.9261 against 0.8832; stretch, Cartesian: 0.9859 against 0.5421; bend, Cartesian: 0.5164 against 0.9688.
Fig. 1 For NH₃’s degenerate stretch and bend, the share of a unit distortion that goes into the E vibration its label names, under three definitions of the distortion, for the published field and for its E partner.

The first result is exact and was not expected. Under the published field, a least-cost distortion of the N–H bonds puts 0.9261 of itself into the degenerate stretch and a one-coordinate distortion 0.8832; for the bond angle the numbers are 0.8832 and 0.9261. Under the E partner, the least-cost stretch is 0.8832 and the one-coordinate stretch 0.9261, and the bond angle the same the other way round. The partner’s least-cost verdicts are the published field’s one-coordinate verdicts, and its one-coordinate verdicts are the published field’s least-cost ones, equal to six decimals.

That has a short explanation once both facts are on the table. Three ways to stretch a bond found that within a two-by-two block the one-coordinate displacement of one coordinate and the least-cost displacement of the other read the same share, because the inverse of a two-by-two matrix is its adjugate: the entries swap across the diagonal. The partner field is itself a swap across the block’s diagonal, rescaled by the kinetic ratio. Two swaps compose into an exchange of definitions, and the shares follow. Only the B-matrix row, which is not one of the pair, comes out with genuinely new numbers: 0.542 for the stretch against 0.986, and 0.969 for the bend against 0.516.

The exchange is exact only for the partner built at NH₃’s own masses. The E partner built with ND₃’s kinetic ratio instead, and then read at NH₃’s masses, gives least-cost shares of 0.8882 for the stretch and 0.9301 for the bend, close to the exchanged values but off them by half a hundredth. That is the same fact seen from the verdicts’ side as from the band centres’: a partner is a partner at one set of masses, and the further an isotope moves the kinetic ratio, the further its partner’s verdicts drift from an exact exchange. For ammonia’s degenerate block deuterium moves that ratio by under two per cent, which is why the drift is small.

The B-matrix row behaves differently because it is not one of the pair the adjugate identity links. Its shares under the partner are not the published field’s numbers in another place; they are new, and they move a long way, the stretch from 0.986 to 0.542 and the bend from 0.516 to 0.969. Under the published field the B-matrix row says a bond stretch is almost all degenerate stretch and a bond-angle opening half degenerate bend; under the partner it says the reverse. A label’s verdict under that definition is therefore the most field-dependent of the three, which matches the earlier finding that the B-matrix row moves a central atom as far as a ligand and so measures as much rigid motion as deformation.

The consequence is a verdict that flips. Nine tenths is the line at which a label settles a coordinate. Under the published field the least-cost bond-angle verdict is 0.883, unsettled; under its partner it is 0.926, settled. Both fields reproduce NH₃’s bands exactly. Which side of the line ammonia’s bond angle falls on depends on which of two spectrally identical fields is the molecule’s.

Off the end of the band

Along the band the bond-angle verdict and ζ₄ move together; the partner is off the end of that line. The share of a least-cost bond-angle distortion that goes into the degenerate bend, against NH₃'s ζ₄, for every field the walk computed, with the band the frequencies allow marked, and for the E partner: 0.926 at ζ₄ = −0.518.
Fig. 2 The least-cost bond-angle verdict against NH₃’s ζ4\zeta_4 for every field on the walk along the stretch–bend constant, and for the E partner.

Along the walked band, the bond-angle verdict and ζ4\zeta_4 move together, almost on a straight line: the verdict crosses nine tenths at ζ4\zeta_4 = −0.696, and within the band the two run from 0.840 at ζ4\zeta_4 = −0.807 to 0.923 at −0.641. The E partner sits at a verdict of 0.926 and a ζ4\zeta_4 of −0.518. It is above the whole band in verdict and beyond its end in ζ4\zeta_4 by 0.12, and it is not on the band’s line: extrapolated, the band’s line would put a verdict of 0.926 near ζ4\zeta_4 = −0.63, not −0.52. The partner is a different kind of field, and the verdict–ζ relation that held along the walk does not carry to it.

ζ₄ for NH₃ under the published field and its partners, against the band the frequencies allow. The first-order Coriolis constant of NH₃'s degenerate bend. Published field −0.732; its E partner −0.518; the E partner built at ND₃'s masses −0.506; the A₁ partner −0.732, the same as the published field's. The shaded band is the range −0.807 to −0.641 that fields the frequencies cannot tell apart give, and the dashed line the ζ₄ of −0.696 at which the bond-angle label's verdict crosses nine tenths.
Fig. 3 NH₃’s ζ4\zeta_4 under the published field and its partners, against the band the frequencies allow and the ζ4\zeta_4 at which the bond-angle verdict crosses nine tenths.

On ζ4\zeta_4 alone the picture is simple. The published field’s −0.732 sits inside the band; the E partner’s −0.518, and the −0.506 of the E partner built at ND₃’s masses instead, both sit well outside it. The gap between the published field and its partner, 0.214, is larger than the whole band’s width of 0.166. A ζ4\zeta_4 known to the ±0.005 that the earlier essay asked of a measurement would separate them forty times over.

What the partner does to the hydrogens

A first-order Coriolis constant is a sum over atoms, each atom contributing the area its displacement ellipse sweeps in the plane perpendicular to the axis, signed by the sense of rotation. For NH₃’s degenerate bend under the published field the nitrogen contributes +0.021 and each hydrogen −0.251. Under the E partner the nitrogen contributes +0.037 and each hydrogen −0.185. Almost all of the change from −0.732 to −0.518 is in the hydrogens.

It is not in how much they move. Each hydrogen carries 0.315 of the bend’s mass-weighted motion in that plane under the published field and 0.301 under the partner, a change of a twentieth. What changes is the shape of its path: under the partner each hydrogen’s ellipse is flatter, so it sweeps less area for nearly the same amount of motion. That is the same mechanism the earlier essay found along the walked band, where the hydrogens’ share stayed between 30.9 and 32.0 per cent while their roundness changed, and it is why ζ reads mode shapes that band centres cannot: a band centre is a sum of motion weighted by stiffness, and a flatter ellipse with the same total motion leaves it unchanged.

The degenerate stretch moves the other way, from +0.509 to +0.296, again mostly through the hydrogens, whose contributions fall from 0.131 each to 0.065 while their share of the motion rises slightly, from 0.231 to 0.246. The two changes are not independent. The sum rule fixes their total, and the hydrogens’ loss in the bend is the stretch’s gain in the opposite sense.

The sum rule still holds, and the partner slides along it

The partners stay on the sum rule's line and move along it, further than the whole band. ζ₃ against ζ₄ for NH₃ and ND₃: the fields of the band the frequencies allow, and the published field's partners. Every point lies on its isotopologue's line ζ₃ + ζ₄ = −0.222 for NH₃ and −0.196 for ND₃. The E partner sits at ζ₄ = −0.518 for NH₃ and −0.579 for ND₃, beyond the band's −0.641.
Fig. 4 ζ3\zeta_3 against ζ4\zeta_4 for NH₃ and ND₃: the fields of the band the frequencies allow, and the E partners, with each isotopologue’s sum-rule line.

The sum rule is not a property of the band. It holds because the degenerate displacements of a symmetric top form a complete space — the sideways translations, the rotations about the two perpendicular axes and the two degenerate vibrations — whose total ζ is exactly one whatever the field. So every field, partners included, must give ζ3+ζ4=I∥/2I⊥−1\zeta_3+\zeta_4 = I_\parallel/2I_\perp - 1, which is −0.2225 for NH₃ and −0.1955 for ND₃, and every one does to a part in a million. The partners slide along the same line as the band, only further: the E partner’s NH₃ point is at ζ3\zeta_3 = 0.296, ζ4\zeta_4 = −0.518; its ND₃ point at 0.384 and −0.579.

That makes ζ3\zeta_3 no help and no hindrance. Measuring both constants adds nothing to measuring one, exactly as before. What the partner adds is reach: the one number the two constants carry can now distinguish fields much further apart than the band, and the most consequential of those is the one the frequencies cannot exclude.

Each measurement is blind where the other sees

Two partners, two measurements, and each measurement blind to the partner the other sees. For the E and A₁ partners of ammonia's published field at NH₃'s masses: how far each misses ND₃'s bands in its block against the published field, its ζ₄ against the published field's, and whether it changes the degenerate bend's verdict on the bond angle.
Fig. 5 For the E and A1A_1 partners: how far each misses ND₃’s bands against the published field, its ζ4\zeta_4 against the published field’s, and whether it moves the degenerate bend’s verdict.

The A1A_1 partner is the mirror case. It exchanges the symmetric stretch and umbrella combinations and leaves the degenerate block untouched, so both Coriolis constants are unchanged to six decimals — a first-order ζ is built from the degenerate vibrations’ shapes and nothing else — and so is every E-block verdict. No measurement of ζ can see it. ND₃’s bands rule it out at once, missing by 4.1 per cent against the published field’s 0.36.

The same holds at ND₃’s masses, where ζ4\zeta_4 is −0.770 under the published field and −0.579 under the E partner built at NH₃’s masses, a gap of 0.19, so a deuterated measurement would separate the two fields almost as well as a normal one. Inside its own block the A1A_1 partner does exactly what the E partner does in the degenerate one. Under the published field a least-cost distortion of the bonds puts 0.882 of itself into the symmetric stretch and a one-coordinate distortion 0.834; for the bond angle the shares into the umbrella are 0.834 and 0.882. Under the A1A_1 partner the four numbers trade places to six decimals, and the B-matrix row again gives new numbers, 0.685 for the stretch against 0.958 and 0.989 for the umbrella against 0.618. So the exchange of definitions is not a property of the degenerate species; it is a property of any two-by-two block and its partner. Every symmetry species that appears twice in a molecule carries the same pair of readings, and a measurement that cannot reach a block’s mode shapes cannot say which reading belongs to the molecule.

So the two kinds of evidence divide the work exactly. Deuterium’s band centres reject the A1A_1 partner and only disfavour the E partner. The Coriolis constant cannot see the A1A_1 partner and separates the E partner by more than the whole band. Neither alone narrows ammonia’s harmonic field to one; the two together do, at least among these four fields.

Each field: its Coriolis constants at both masses and its least-cost verdicts in both blocks. For the published field and its three partners: ζ₃ and ζ₄ for NH₃, ζ₄ for ND₃, and the least-cost concentration of the degenerate stretch, the degenerate bend and the umbrella.
Fig. 6 Each field’s Coriolis constants at both masses and its least-cost verdicts for the degenerate stretch, the degenerate bend and the umbrella.

How the verdicts and constants were read

The fields are the five-constant valence field fitted to NH₃ and ND₃ and its partners from the previous construction: each block’s stretch and bend combinations exchanged with the block’s own ratio of kinetic elements, read from the normal-mode calculation, with the stretch–bend constant and the other block held. The concentrations are those of the three-displacement census, computed for each field from its own mode shapes. The Coriolis constants are first-order: for each degenerate pair, the sum over atoms of the cross products of the two partners’ mass-weighted displacements in the plane perpendicular to the axis, with the pair oriented so that the three-fold rotation turns the first towards the second. The band and the verdict’s crossing are the walk along the stretch–bend constant, refitting the other four constants to both molecules at each step.

Three things are checked where the figures are drawn: the sum rule holds on every field at both isotopologues to 10−610^{-6}; the E partner’s least-cost and one-coordinate concentrations equal the published field’s one-coordinate and least-cost ones to 10−610^{-6} for both coordinates; and the A1A_1 partner leaves both Coriolis constants unchanged to 10−610^{-6}. The refusal is that the A1A_1 partner must change the A1A_1 block’s verdicts, by more than a hundredth: a partner that moved nothing anywhere would mean the construction had rebuilt the published field. It moves the umbrella’s least-cost verdict from 0.834 to 0.882.

A harmonic constant is not a band analysis

The ζ here is harmonic. The earlier essay spelled out what stands between a band analysis’s effective ζ and the harmonic one: anharmonic contributions, the resonance of the degenerate bend with the umbrella’s overtone, and the fact that every level of ammonia is a tunnelling doublet. None of that has changed. What has changed is the size of the target: separating the published field from its E partner needs ζ4\zeta_4 to about ±0.1, not ±0.005, and corrections of a few hundredths would not blur it.

Four fields, not all fields. The partner construction gives one exact partner per block at one set of masses, and the walked band is a one-parameter family fitted to both molecules. The fields consistent with ammonia’s data to within a harmonic model’s noise form a larger set than either, and the four computed here are landmarks in it rather than the whole of it.

The verdict’s line is a convention. Nine tenths is the census’s threshold for “settled”, chosen before any of this; a verdict of 0.883 against 0.926 is a large difference by any threshold, but whether it is called settled is a choice.

A label’s verdict is a statement about a pair of fields

The deeper point is about what a symmetry label’s verdict is. Ammonia’s E label claims its bond angle; whether the claim holds to nine tenths was found to depend on the definition of the distortion, then on where in the band the field sat. It now turns out that the two dependences are one: the field’s partner reads the other definition. A molecule whose bands determine its field only up to a block exchange has, in effect, a verdict that is a pair of numbers, one per definition, and which one is “the” least-cost verdict is decided by which of the pair is the molecule’s field. The frequencies do not say. The Coriolis constant does.

There is a practical form of that. A census of symmetry-label verdicts computed from a fitted force field should report, for every species that appears twice, both the least-cost and the one-coordinate share, because the partner field guarantees that the pair is the same pair whichever field is right and only their assignment to the two definitions is in doubt. For ammonia’s bond angle the pair is 0.883 and 0.926, and the honest statement is that the label settles one definition of the distortion and not the other, with the Coriolis constant as the measurement that would say which. That is a weaker claim than a single verdict and a much more durable one.

Still open: the measured constants, and a molecule where the exchange is not a choice

The obvious open question is the data step that has been open since the Coriolis constants were first computed. Ammonia’s ν4\nu_4 and ν3\nu_3 band analyses report effective ζ, and the separation needed here, a tenth or more, is far coarser than the one the band required. Reading the published effective constants, even with corrections only estimated, would probably say which of the two fields ammonia’s bend belongs to, and so on which side of nine tenths its bond-angle verdict falls.

The nearer question is a pyramid in which the exchange is forced rather than chosen. The partner exists because the block is two by two; a molecule whose degenerate species appears three times has a three-by-three block, where the rescaled exchange of two diagonal elements no longer preserves both invariants and a partner, if any exists, has to be found rather than written down. Whether such a block still trades the definitions of a distortion, or whether the coincidence found here is a property of two-by-two blocks alone, is the next calculation.

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.

Named objects

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

Force constantIrreducible representationsIsotope substitutionModel limitNormal modeValence force field