What all that machinery was for
By now you can walk up to a molecule, hunt down its symmetry elements, assign its point group, and read the columns of a character table without flinching. Earlier in this rung you also learned to take a messy collection of objects — the stretches of every bond, or the orbitals on every ligand — treat it as a [[reducible-representation|reducible representation]], and break it down into the clean [[irreducible-representation|irreducible representations]] listed in the table. That reduction felt like an abstract finger exercise. This guide is where it pays for itself.
Two payoffs are waiting. First, spectroscopy: symmetry tells you, before you ever touch an instrument, which molecular vibrations will absorb infrared light and which will scatter in Raman — and famously, when a band will appear in one technique but not the other. Second, bonding: symmetry tells you how to glue the orbitals scattered across several ligands into a few well-behaved bundles that a central atom's orbitals can actually grab onto. Both rest on the same single idea, so let us state it plainly.
Infrared and Raman: the same molecule, two questions
A molecule vibrates in a fixed set of patterns called normal modes — a symmetric stretch where all bonds breathe in and out together, a bend where the molecule flexes, and so on. Each mode, it turns out, carries a symmetry label: when you run the reduction you learned earlier, every normal mode comes out belonging to one irreducible representation of the molecule's point group. That label is the whole story, because the two great vibrational techniques each ask a symmetry question about it.
Infrared absorption happens when a vibration makes the molecule's dipole moment wobble — light's electric field can only push on a vibration that shifts charge back and forth. So a mode is infrared-active exactly when it changes the dipole, and the dipole points along x, y, or z. Here is the gift the character table hands you: those very letters, x, y and z, are printed in the right-hand column next to the irreducible representations they transform as. A normal mode is IR-active if and only if its label matches the label of x, y, or z. You read it straight off the page.
Raman scattering asks a subtler question. Instead of the dipole, it watches the molecule's polarizability — how squishy its electron cloud is, how easily an external field distorts it. A mode is Raman-active when the vibration changes that squishiness. Polarizability transforms like the quadratic functions x^2, y^2, z^2, xy, xz, yz — and these too are printed in the character table, in the far-right column. So the recipe mirrors infrared exactly: a mode is Raman-active if and only if its label matches the label of one of those quadratic functions. Same table, second column over.
The rule of mutual exclusion, and why CO2 betrays its shape
Now a beautiful consequence falls out for free. In any molecule that has a center of inversion — a point through which every atom has an identical twin on the far side — the linear functions x, y, z and the quadratic functions x^2, xy, and so on can never carry the same symmetry label. The linear ones are always 'ungerade' (sign-flipping through the center), the quadratic ones always 'gerade' (sign-keeping). So no single vibration can change both the dipole and the polarizability. This is the rule of mutual exclusion: in a centrosymmetric molecule, a vibration is IR-active or Raman-active, never both.
Turn that around and it becomes a tool for figuring out shapes you cannot see. Take carbon dioxide. Is it bent (like water) or linear? A bent O-C-O would have no center of inversion, so its symmetric stretch could be both IR- and Raman-active. A linear O=C=O does have a center of inversion right on the carbon. Run the experiment: CO2's symmetric stretch shows up in Raman but is silent in the infrared, while its asymmetric stretch and bend appear in the infrared but not in Raman. No band is shared. Mutual exclusion holds — so the molecule must be linear and centrosymmetric. Spectroscopy plus symmetry settled the geometry without a single diffraction photograph.
CO2, linear, has a center of inversion (point group D-infinity-h): symmetric stretch O<-- C -->O dipole unchanged -> Raman active, IR SILENT asymmetric stretch O--> C--> O dipole wobbles -> IR active, Raman silent bend (x2) O C O (flex) dipole wobbles -> IR active, Raman silent No band appears in BOTH -> mutual exclusion -> the molecule is centrosymmetric (linear). Count check: 3 atoms -> 3N - 5 = 4 vibrational modes (the two bends are a degenerate pair).
SALCs: bundling ligand orbitals so the metal can grab them
Now the bonding payoff, the one the chapters ahead lean on hardest. In an octahedral complex like [Co(NH3)6]3+, six ligands each point a lone pair straight at the central metal. You might be tempted to draw six separate metal-ligand bonds and stop. But the central atom does not offer six identical orbitals to bond with — it offers an s, three p's, and a set of d's, each with its own shape and symmetry. The honest question is: how do you combine the six scattered ligand orbitals into bundles whose symmetries match the metal's orbitals one-for-one? Those tailored bundles are [[symmetry-adapted-linear-combinations|symmetry-adapted linear combinations]], or SALCs.
The procedure is exactly the reduction you already practised, now pointed at the ligands. You treat the set of six ligand donor orbitals as a reducible representation, find its character under each symmetry operation by counting how many orbitals stay put, and reduce it into irreducible representations. For the six sigma-donors of an octahedron the result is a tidy package: a1g + eg + t1u. Each of those labels names a SALC — a specific in-phase or out-of-phase blend of the six ligand orbitals — and each is built to overlap with a metal orbital that wears the same label.
- Pick the orbitals you want to bundle (here: the six sigma-donor lone pairs pointing at the metal) and treat the whole set as one reducible representation.
- For each symmetry operation of the point group, count how many ligand orbitals are left unmoved — that count is the character. (An orbital that gets swapped to another position contributes zero.)
- Use the reduction formula against the character table to split that representation into irreducible pieces — for octahedral sigma-donors you get a1g + eg + t1u.
- Match each ligand SALC to the central-atom orbital carrying the same label, let them combine, and you have the bonding and antibonding levels of the molecular-orbital diagram.
From SALCs to the ligand-field diagram you already half-know
Watch what the matching does. The metal s orbital is totally symmetric — it carries the a1g label — so it pairs with the a1g SALC. The three metal p orbitals transform together as t1u, so they pair with the t1u SALC. Among the five d orbitals, the two that point straight at the ligands (the dz2 and dx2-y2 pair) transform as eg and pair with the eg SALC. That leaves the other three d orbitals — dxy, dxz, dyz, which point into the gaps between ligands — labelled t2g, with no ligand SALC of matching symmetry to combine with. By symmetry alone, those three are forced to stay nonbonding.
And there, dropping out of pure symmetry, is the famous picture. The eg pair (dz2, dx2-y2) is pushed up because it forms antibonding combinations with the ligands aimed straight at it; the t2g trio (dxy, dxz, dyz) sits lower, untouched and nonbonding. The gap between them is the splitting you will soon call delta-o. This is the same five-d-into-two-sets result that simple [[crystal-field-theory|crystal field theory]] gets by imagining ligands as point charges — but ligand field theory, built on SALCs and real orbital overlap, earns it honestly and explains things crystal field theory cannot, such as why pi-accepting ligands enlarge the gap.
Be clear about what symmetry does and does not give you. The labels and the splitting pattern (which orbitals group together, which stay nonbonding) come from symmetry alone and are exact. But symmetry never tells you the size of delta-o, nor whether a complex ends up high-spin or low-spin — that depends on the actual energies and the electron pairing energy, which you measure or compute. Symmetry draws the staircase; it does not tell you how tall each step is.
Selection rules are just symmetry in disguise
Pull back and the same idea governs why complexes are coloured. An electronic transition — say an electron hopping between two of those d levels, a d-d transition — is allowed only if the symmetries line up, this time of the starting orbital, the light's electric field, and the ending orbital. The famous [[laporte-selection-rule|Laporte rule]] is precisely this: in a centrosymmetric molecule, the light operator is ungerade, but two d orbitals are both gerade, so the symmetry product comes out gerade — a mismatch — and the transition is forbidden. It is the very same gerade/ungerade bookkeeping that gave you mutual exclusion in vibrations.
But many octahedral complexes are visibly, gorgeously coloured — so the rule must leak. It does, gently and honestly. The molecule is always vibrating, and certain vibrations momentarily destroy the center of inversion; for that fleeting instant the gerade/ungerade labels stop applying and the transition borrows a little allowedness. This loophole, called vibronic coupling, is exactly why a Laporte-forbidden d-d band is weak and pale rather than truly black. The rule is real, the colour is faint, and the faintness is the rule half-broken — a far more honest story than 'd-d transitions are simply allowed.'
Hold the whole picture for a moment. The same one-line idea — symmetries must match or the integral vanishes — told you which vibrations show in infrared versus Raman, which orbital combinations are allowed to bond, and which electronic transitions can absorb light. Earlier rungs in this ladder showed you symmetry as a way to classify and to spot chirality; this rung has turned it into a working predictive engine. In the chapters ahead — coordination chemistry, ligand-field theory, the d-d and charge-transfer spectra of complexes — you will not be learning these rules afresh. You will be watching the symmetry you already command quietly do the heavy lifting.