What symmetry decides

No tetrahedral isotope can choose methane's field

Methane's published force field misses its lines by 0.401 per cent, and only 0.037 of that is forced by the product rule. Untie the E bend from the T₂ bend — the second bend–bend constant the redundancy had removed — and the field becomes exactly determined by CH₄ and CD₄, with the whole misfit gone down to the floor. But the T₂ block is a quadratic and has two roots, F₃₄ = 0.23 and 1.17 mdyn/Å. The Coriolis constant proposed to separate them is the same for both to eight decimal places, and so is every ¹³C shift: a reflection that preserves both atoms' kinetic matrices carries one root onto the other at every mass. Only a molecule that is no longer a tetrahedron can tell them apart, and CH₃D puts their C–D stretches 519 wavenumbers apart.

Worth reading first: The theorem uses up an equation per species · Two Coriolis constants are one number.

Methane’s force field in this collection has four constants: the C–H stretch, the H–C–H bend, an interaction between two stretches and an interaction between a stretch and a bend it is a leg of. It reproduces the eighteen lines of CH₄ and CD₄ to 0.401 per cent. The theorem uses up an equation per species found that almost none of that misfit is forced. The Teller–Redlich product rule, which fixes one combination of each symmetry species’ lines for every harmonic field, accounts for 0.037 per cent of it, 0.8 per cent of the sum of squares, and the rest sits in two places the rule cannot see. In the T2T_2 species the field puts too large an isotope shift on the bend and too small a one on the stretch. In E it has the frequency scale slightly low in both molecules.

Both errors have the same suspect. The bend–bend constant was left out because the six angles at a tetrahedral centre are not independent: five bending coordinates span every deformation, so the bend constant and a single bend–bend constant can run together along a direction no frequency sees, and when allowed to they ran off to −97,596 mdyn/Å each. Dropping one cost nothing then. That essay’s continuation asked what restoring it under a constraint would do. It would use methane’s last spare equation, and the field would become exactly determined, as water’s is.

It does, and the misfit goes. But an exactly determined methane field turns out to be two fields, and the two measurements that look as if they should choose between them cannot. The reason is a symmetry of the masses, not of the molecule.

Two bend–bend constants, and the tie they undo

The valence form ties two things together that the molecule does not. In symmetry coordinates — combinations of the four stretches and six angles that each belong to one species — methane’s general harmonic field has five constants: F11F_{11} for the A1A_1 stretch, F22F_{22} for the E bend, and a two-by-two block F33F_{33}, F34F_{34}, F44F_{44} for the T2T_2 species, where a stretch and a bend of the same symmetry mix. The valence constants map onto these, and with no bend–bend constant, or with one shared by every pair of angles, the map gives F22=F44F_{22} = F_{44}: the E bend and the T2T_2 bend have the same stiffness, whatever the bend constant is.

That tie is what the published field carries, and it explains both of its errors at once. The E lines want one bend constant and the T2T_2 bend wants another. A field forced to give them the same one compromises, and it settles the compromise in T2T_2 by moving the stretch–bend interaction, which is the knob that decides how the T2T_2 isotope shift divides between the two modes.

Two bend–bend constants untie it — one, fθθf_{\theta\theta}, for angles that share a bond and one, fθθ′f'_{\theta\theta}, for the three pairs of opposite angles that share none. Then F22=fθ−2fθθ+fθθ′F_{22} = f_\theta - 2f_{\theta\theta} + f'_{\theta\theta} and F44=fθ−fθθ′F_{44} = f_\theta - f'_{\theta\theta}, and raising all three together changes neither, which is the redundancy appearing as a direction that costs nothing rather than as a runaway. Only the difference fθθ−fθθ′f_{\theta\theta} - f'_{\theta\theta} is new, so restoring “the” bend–bend constant without the runaway means restoring two and keeping one combination. The count then matches the lines exactly. A1A_1 and E each have one line in each molecule, and the product rule fixes their ratio, so each species fixes its one constant once. T2T_2 has four lines, the rule fixes one combination, and three are left for three constants.

Exactly determined, and twice

Every T₂ block that gives CH₄'s two lines exactly is one circle, and CD₄'s ratio of lines crosses its observed value at two points. Along the one-parameter family of T₂ force-constant blocks that reproduce CH₄'s 1,367.4 and 3,156.8 cm⁻¹ exactly, the ratio of CD₄'s two calculated T₂ lines, from 2.064 to 2.582, against the observed 2.2551. It crosses twice: at angle 1.535 (F₃₃ 5.390, F₃₄ 0.227, F₄₄ 0.467 mdyn/Å) and at 2.814 (F₃₃ 1.630, F₃₄ 1.167, F₄₄ 2.346).
Fig. 1 Along the family of T2T_2 blocks that give CH₄’s two T2T_2 lines exactly, the ratio of CD₄’s two calculated T2T_2 lines; it crosses the observed ratio at two blocks.

The T2T_2 block is solved in two steps. A two-by-two block that reproduces CH₄’s two T2T_2 lines exactly is fixed by a trace and a determinant, which leaves one free angle. Every such block can be written as the kinetic matrix’s inverse square root, times a rotation of the two eigenvalues, times the inverse square root again. The angle runs once round a circle, and as it does, CD₄’s two calculated lines move. Their product never moves, because that is the product rule, so all that changes along the circle is their ratio, which swings between 2.064 and 2.582. The observed ratio is 2.2551. A closed curve crosses a level twice or not at all, and here it crosses twice.

Each crossing, refined against all four T2T_2 lines by least squares, is a field. The first has F33=5.390F_{33} = 5.390, F34=0.227F_{34} = 0.227 and F44=0.467F_{44} = 0.467 mdyn/Å, with bends scaled by the bond length. The second has F33=1.630F_{33} = 1.630, F34=1.167F_{34} = 1.167 and F44=2.346F_{44} = 2.346. With F11=5.440F_{11} = 5.440 and F22=0.4957F_{22} = 0.4957 from the one-line species, each is a complete harmonic field for methane, and the two reproduce CH₄’s and CD₄’s eight distinct lines identically.

Untied, the field misses each line only by its share of the product rule's floor: the T₂ division and the E scale both come right. Each observed line of CH₄ and CD₄ with its relative misfit under the published four-constant field (upper bar) and under the first exactly determined general field (lower bar; the second gives the same lines): CH₄ A₁ symmetric stretch 3025.5: 0.040% against 0.040%; CH₄ E bend 1582.7: -0.168% against -0.012%; CH₄ T₂ bend 1367.4: -0.340% against 0.041%; CH₄ T₂ stretch 3156.8: 0.522% against 0.041%; CD₄ A₁ symmetric stretch 2141.9: -0.040% against -0.040%; CD₄ E bend 1119.3: -0.144% against 0.012%; CD₄ T₂ bend 1035.9: 0.529% against -0.041%; CD₄ T₂ stretch 2336.1: -0.513% against -0.041%. Residuals 0.401% and 0.037%; the theorem's floor is 0.037%.
Fig. 2 Each line of CH₄ and CD₄ with its relative misfit under the published field and under the first exact field.

What they reproduce them to is the floor. The residual of either exact field over all eighteen lines is 0.0369 per cent, equal to the product rule’s floor for methane to 10−910^{-9}, because an exactly determined field has nothing left to trade: each species’ lines are matched except for the share of its observed isotope product that disagrees with the theorem. The published field’s T2T_2 misfits of half a per cent become 0.042, the same size on all four lines with the signs the theorem dictates. Its E lines, 0.17 per cent low in CH₄ and 0.14 in CD₄, become 0.01 low and 0.01 high. So the question the equation count left open has an answer. Methane’s misfit was the missing constant, and the valence form with its tie was the whole of the error. The tie itself is small: the exact fields put the E bend 6.2 per cent stiffer than the T2T_2 bend, fθθ−fθθ′=−0.015f_{\theta\theta} - f'_{\theta\theta} = -0.015 mdyn/Å.

Three fields in symmetry constants: the published one ties E to the T₂ bend; both exact roots untie it, and the second is no C–H bond. Symmetry force constants in mdyn/Å (bends scaled by the bond length), the stretch–stretch interaction fᵣᵣ they imply, and the residual over all eighteen lines: published (four constants): F₁₁ 5.440, F₂₂ 0.4941, F₃₃ 5.478, F₃₄ 0.489, F₄₄ 0.4941, fᵣᵣ -0.010, residual 0.401%; first exact root: F₁₁ 5.440, F₂₂ 0.4957, F₃₃ 5.390, F₃₄ 0.227, F₄₄ 0.4666, fᵣᵣ 0.013, residual 0.037%; second exact root: F₁₁ 5.440, F₂₂ 0.4957, F₃₃ 1.630, F₃₄ 1.167, F₄₄ 2.3462, fᵣᵣ 0.952, residual 0.037%.
Fig. 3 The published field and the two exact roots in symmetry force constants, with the stretch–stretch interaction each implies and its residual.

One root is a C–H bond and one is not

The table shows the difference between the roots in the units a chemist reads. The first root’s T2T_2 stretch constant, 5.390, sits beside an A1A_1 constant of 5.440. The two differ by four times the stretch–stretch interaction, which comes out at 0.0125 mdyn/Å, a coupling of a quarter of a per cent between two C–H bonds on one carbon. That is the size every valence field of methane has found. The second root’s T2T_2 stretch constant is 1.630, so the same arithmetic gives a stretch–stretch interaction of 0.952, nearly a fifth of the symmetric stretch constant and more than a third of the 2.583 it leaves for the bond itself. It would make the antisymmetric stretch of methane three times softer than the symmetric one. Its T2T_2 bend constant, 2.346, is nearly five times the E bend’s.

What the T₂ modes are made of: under the first root each is nearly pure, under the second the 'stretch' is mostly bend. Each T₂ mode's potential energy divided between the stretch coordinate, the bend coordinate and their cross term, under each field: published (four constants), bend: stretch 0.096, bend 1.097, cross -0.193; published (four constants), stretch: stretch 1.001, bend 0.000, cross -0.001; first exact root, bend: stretch 0.011, bend 1.019, cross -0.030; first exact root, stretch: stretch 1.010, bend 0.002, cross -0.011; second exact root, bend: stretch 1.535, bend 0.718, cross -1.253; second exact root, stretch: stretch 0.018, bend 0.835, cross 0.147.
Fig. 4 Each T2T_2 mode’s potential energy divided between stretch, bend and cross term, under the published field and each exact root.

The mode shapes say the same thing in another way. Under the first root the T2T_2 modes are as pure as normal modes get: the bend carries 1.019 of its potential energy in the bend coordinate and −0.030 in the cross term, the stretch 1.010 in the stretch. The published field, by contrast, gives its bend a tenth of a stretch and a fifth of an opposed cross term, which is its compromise made visible. Under the second root the lower mode spreads its energy over stretch and bend with a large opposed cross term, 1.535, 0.718 and −1.253, and the upper mode, the one at 3,158 cm⁻¹, holds 0.835 of its energy in the bend. A line at the C–H stretching frequency would be a bend.

Chemistry rejects the second root at once. The point is that nothing in methane’s spectrum does. Both roots reproduce every line of CH₄ and CD₄ to the same floor, and a judgement that one of them is “physical” is a judgement brought from other molecules. The census of force fields has been reading symmetry labels’ verdicts off fields that fit their lines, and here is a molecule where the lines fit two fields, one of which describes no molecule at all. The obvious measurement to settle it is the one the ammonia essays kept returning to.

The Coriolis constant was already measured

A degenerate vibration in a rotating molecule carries angular momentum, and the Coriolis constant ζ says how much: it is read from the vibration’s shape, not its frequency, and it shows up in a rotationally resolved band as a splitting no harmonic frequency predicts. For a tetrahedral molecule the two T2T_2 constants obey a sum rule, ζ3+ζ4=12\zeta_3 + \zeta_4 = \tfrac12, whatever the field, which is the tetrahedral case of the identity that made ammonia’s two constants one number. So methane has one Coriolis number to offer, and the two roots have very different mode shapes, so they should give very different values of it.

Every field exact for CH₄ puts ζ₃ on one curve of CD₄'s ratio: the two roots, with the same ratio, have the same Coriolis constant. For each T₂ block that reproduces CH₄'s two T₂ lines, the Coriolis constant ζ₃ of the stretch against the ratio of CD₄'s two calculated T₂ lines, coloured by whether the block's stretch constant exceeds its bend constant. Both branches lie on one curve. At the observed ratio 2.2551 both roots give ζ₃ = 0.0347 (to 10⁻⁸ of each other); the published field, exact for neither isotopologue, gives -0.0861. ζ₃ + ζ₄ = ½ for every field.
Fig. 5 For every T2T_2 block that reproduces CH₄’s two T2T_2 lines, the Coriolis constant ζ3\zeta_3 of the stretch against the ratio of CD₄’s two calculated T2T_2 lines, coloured by branch.

They give the same value: ζ3=0.034728\zeta_3 = 0.034728 for both, equal to eight decimal places. The figure shows why it is not an accident of these two fields. Along the whole circle of blocks that reproduce CH₄, the Coriolis constant is a single function of one number, CD₄’s ratio of lines. Blocks where the stretch constant leads and blocks where the bend leads fall on one curve. Any two fields that agree on CH₄ and CD₄ agree on ζ3\zeta_3, so a measured Coriolis constant can only confirm what the isotopologue already said or contradict it, and cannot choose between fields the isotopologue allows. That is the same kind of thing the product rule does to an isotopologue’s lines: a quantity measured as if it were new is a function of quantities already in hand.

The published field, fitted to neither molecule exactly, gives ζ3=−0.086\zeta_3 = -0.086, the opposite sign to the value both exact fields share. So the four-constant form’s misfit, small in frequency, is large in the one number methane’s band analyses report. Here, then, is a property of the published field that a frequency residual of 0.4 per cent did not reveal.

A reflection that every isotope keeps

The reason runs through the kinetic energy. In the T2T_2 species the kinetic matrix GG is a sum of two parts, one multiplied by the inverse hydrogen mass and one by the inverse carbon mass. The hydrogens’ part is a full two-by-two matrix. The carbon’s is rank one, because in each T2T_2 component the carbon moves in one direction only, against the four hydrogens, so its part has the form b bTb\,b^{\mathsf T} for a single vector bb. Write the hydrogens’ part as WWTWW^{\mathsf T}. Then the reflection that fixes the direction W−1bW^{-1}b and reverses the one perpendicular to it, carried back through WW, is a linear map TT that leaves both parts unchanged. So TGTT=GTGT^{\mathsf T} = G for every hydrogen mass and every carbon mass together.

A map that preserves GG maps fields to fields with the same frequencies: F′=T−TFT−1F' = T^{-\mathsf T} F T^{-1} gives GF′GF' the same eigenvalues as GFGF. The second root is exactly the first carried through TT, to 10−710^{-7} in each constant. The mirror is a property of the masses’ geometry and has nothing to do with the field. Since it holds at every mass, no substitution of hydrogen or carbon that keeps the tetrahedron can tell the two fields apart. The Coriolis constant agrees because it is built from the same kinetic parts.

Carbon-13 cannot choose either: both exact roots predict the same ¹³CH₄ shifts, to a billionth of a wavenumber. The ¹²CH₄ − ¹³CH₄ shift of each T₂ line under each field, in cm⁻¹; the A₁ and E lines do not move, since carbon is still in those modes: published (four constants): bend 9.670, stretch 8.480; first exact root: bend 8.641, stretch 10.909; second exact root: bend 8.641, stretch 10.909.
Fig. 6 The ¹²CH₄ − ¹³CH₄ shift of each T2T_2 line under the published field and each exact root.

Carbon-13 is the natural thing to try, since it is the only other stable substitution methane has and it moves the one atom the hydrogens’ data did not. Its shifts are identical under the two roots — 8.640850 cm⁻¹ for the bend and 10.909470 for the stretch, equal to 10−910^{-9} — and the published field gives 9.67 and 8.48, the stretch and bend shifts the other way round. The A1A_1 and E lines do not move at all, since the carbon stands still in them. A tritium substitution, CT₄, would change the hydrogens’ mass and nothing else, and fall to the same mirror. Every isotopologue of methane that is still a regular tetrahedron reproduces the same two fields.

CH₃D breaks the mirror

The mirror needs the kinetic matrix to split into a hydrogen part and a carbon part, each the same in every T2T_2 component, and that needs all four hydrogens to have one mass. Replace one of them and the molecule falls to C3vC_{3v}. The T2T_2 species splits into A1A_1 and E, the stretch and bend blocks mix with the former A1A_1 and E coordinates, and no single reflection preserves the new kinetic matrix.

Partial deuteration breaks the tetrahedron and the mirror with it: the two roots put CH₃D's lines hundreds of wavenumbers apart. Every harmonic line of CH₃D and CH₂D₂ under each exact root, in cm⁻¹: first root: CH₃D 1206.7, 1206.7, 1363.4, 1521.6, 1521.6, 2281.7, 3065.2, 3157.6, 3157.6; CH₂D₂ 1068.4, 1135.1, 1288.3, 1370.7, 1486.5, 2232.0, 2335.4, 3098.9, 3157.2. second root: CH₃D 1087.2, 1350.4, 1350.4, 1465.3, 1465.3, 2800.8, 2930.1, 2930.1, 3131.5; CH₂D₂ 1037.5, 1141.3, 1297.4, 1348.1, 1370.7, 2561.0, 2658.3, 2896.8, 2900.0.
Fig. 7 Every harmonic line of CH₃D and CH₂D₂ under each exact root.

The two roots then disagree by more than anything else on the page. In CH₃D the first root puts the C–D stretch at 2,281.7 cm⁻¹, between CD₄’s symmetric and antisymmetric stretches, where a lone C–D bond belongs. The second puts it at 2,800.8, far above CD₄’s own C–D stretches at 2,141.9 and 2,336.1, and moves the C–H stretches down to 2,930 and 3,131. The bends disagree by up to 120 cm⁻¹. In CH₂D₂ the two C–D stretches come at 2,232 and 2,335 under the first root and 2,561 and 2,658 under the second. One rotationally assigned band of a partly deuterated methane separates the roots by hundreds of wavenumbers, where carbon-13 separated them by none.

The general lesson is about where information lives. The deuterated molecule that made ammonia’s partner field only twice less likely was fully deuterated, and so was every isotopologue the census of fitted fields has used. Full substitution is convenient because it keeps the symmetry, and with it the species and the block-diagonal secular equation. For the same reason it keeps every symmetry of the kinetic energy, and those are what make the two-root ambiguity invisible to it.

How the fields were found

Symmetry coordinates are built from methane’s ten valence coordinates: the A1A_1 stretch as the normalised sum of the four bonds, each T2T_2 stretch component as the bonds weighted by that Cartesian component of their direction, each T2T_2 bend component as the angles weighted by the same component of the sum of their two bonds’ directions, the redundant A1A_1 bend as the normalised sum of the angles, and the E pair by Gram–Schmidt on what is left. Angles are scaled by the bond length, so all force constants are in mdyn/Å. The kinetic matrix is BM−1BTB M^{-1} B^{\mathsf T} from the valence B matrix, transformed. The one-line species’ constants are least-squares fits in the logarithm of their two lines. The T2T_2 roots are bracketed where CD₄’s ratio crosses the observed one along the CH₄-exact circle, bisected, and refined by Gauss–Newton against the four T2T_2 lines. Coriolis constants are read from the Cartesian normal modes, mass-weighted, as the magnitude of the antisymmetric matrix of the angular-momentum operator within each triply degenerate set.

Each piece is checked against something it did not use. The symmetry coordinates are orthonormal, the redundant bend has zero kinetic energy to 10−1010^{-10}, and the published field transformed into them has no coupling between species. Its symmetry route reproduces the frequencies its own valence route gives, and its E and T2T_2 bend constants are equal, the tie, to 10−1010^{-10}. Each exact field is written back into all ten valence coordinates and diagonalised as a fifteen-dimensional Cartesian problem, a route with nothing in common with the two-by-two algebra, and gives the same frequencies to 10−510^{-5} cm⁻¹. Each residual equals the product rule’s floor to 10−910^{-9}. ζ3+ζ4=12\zeta_3 + \zeta_4 = \tfrac12 holds for every field, and ∣ζ∣|\zeta| is the same about all three axes. The mirror carries the first root onto the second to 10−710^{-7}. And there is a refusal. The published form, with E tied to the T2T_2 bend, must not reach the floor at any point of the CH₄-exact circle, and it cannot: its best residual there is 0.520 per cent. Had it reached the floor, the tie would not be what the misfit was.

What the two roots do not settle

These are harmonic fields fitted to harmonic frequencies. The lines are the harmonic wavenumbers every earlier fit of methane’s field was made to, and every statement here is about the quadratic part of the potential. Real band origins carry anharmonicity that differs between isotopologues, and that difference is where the 0.037 per cent floor comes from; a field fitted to fundamentals would have no mirror-protected degeneracy of this exact kind, but it would have more constants and its own unidentifiable directions.

CH₃D’s numbers are predictions, not comparisons. No CH₃D or CH₂D₂ spectrum is used here. The claim is that the two roots disagree about those molecules by hundreds of wavenumbers, which any assigned band would settle, not which root a band supports — though only one of them puts a C–D stretch where a C–D stretch can be.

The mirror is a statement about T2T_2. It exists because the carbon’s kinetic part is rank one in a two-by-two block. A larger block, or a central atom that moves in more than one direction per component, would need a different argument, and the same kind of degeneracy is not guaranteed elsewhere.

Nothing here touches the labels’ verdicts. The census of symmetry labels read verdicts off the published field. The first root changes the T2T_2 mode shapes from a compromise to nearly pure motions, and the verdicts that depend on them would move. Recomputing them is a separate calculation.

Two fields that agree on everything symmetric

The isotope shift is arithmetic, species by species, and the arithmetic has a symmetry of its own that the molecule does not. Methane’s point group decides which coordinates mix. The masses decide which mixtures are distinguishable, and for a tetrahedral XY4XY_4 they leave one reflection in each T2T_2 block that no substitution of X or Y can break. The surprising thing is not that the quadratic has two roots — quadratics do — but that the two measurements chosen to separate them, the Coriolis constant and the central atom’s isotope, are both blind to the same reflection, because both read the kinetic energy that the reflection preserves. What separates the roots is a measurement that breaks the symmetry. Symmetry factored the secular equation and made each species’ fit small enough to solve exactly, and in doing so it made methane’s field indeterminate to every experiment that keeps the symmetry.

Still open: the labels under the exact field, and a block of three

The obvious open question is the census. Every symmetry label’s verdict on methane was read off a field whose T2T_2 modes are a compromise, and the exact field’s first root makes them nearly pure. Reading the T2T_2 verdicts again under it — and checking that the second root, carried through the mirror, gives the same verdicts or different ones — would say whether a label’s verdict is a property of the spectrum or of a choice the spectrum cannot make.

The nearer question is a species that appears more than twice. The mirror needs the kinetic matrix to be a full part plus a rank-one part, which a two-by-two block of an XY4XY_4 guarantees. A tetrahedral carbonyl such as Ni(CO)₄ has four T2T_2 modes, and its kinetic matrix has three parts — the nickel’s, rank one, and the carbons’ and oxygens’, each fuller. Whether reflections that preserve all three still exist, and so whether a fully substituted carbonyl’s field is ambiguous in the same way, is the next calculation.

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 substitutionMode compositionNormal modeValence force field