Sigma and Pi Molecular Orbitals and Symmetry Labels

The σ/π/g/u vocabulary is not a historical relic — it is the everyday operating language of spectroscopists, computational chemists, and materials scientists. Photoelectron spectroscopy assigns ionization bands of diatomic and small polyatomic molecules directly by symmetry label (removing an electron from $3\sigma_g$ versus $1\pi_u$ in N2 produces spectroscopically distinguishable ionic states with different vibrational progressions, because ionizing a bonding orbital lengthens the bond and produces a longer Franck–Condon vibrational envelope than ionizing a near-nonbonding orbital). Every quantum chemistry software package (Gaussian, ORCA, Q-Chem, Psi4) labels computed canonical molecular orbitals with irreducible representation symbols derived from exactly this g/u, σ/π machinery whenever the molecular point group is recognized as linear or centrosymmetric, and chemists read these labels directly off the output to interpret UV-vis and photoelectron spectra. In organometallic and inorganic chemistry, the δ-bond concept extending this same $m_\ell$ classification to d orbitals explains quadruple metal–metal bonds such as the Re2Cl8²⁻ ion, where one σ, two π, and one δ bond between rhenium centers were identified by F. Albert Cotton using exactly the node-counting and symmetry arguments developed here, opening an entire subfield of multiply-bonded transition-metal cluster chemistry. In materials science, band-structure calculations for one-dimensional conducting polymers and carbon nanotubes classify crystal orbitals along the periodic axis using the same σ/π distinction, with π-conjugated backbones (polyacetylene, graphene sheets, carbon nanotubes) responsible for the delocalized conduction electrons that make these materials useful in organic electronics and flexible displays. In biochemistry, the planarity and restricted rotation of the peptide bond is a direct consequence of π-type delocalization between the carbonyl and the amide nitrogen — while amides are not centrosymmetric so g/u labels do not apply, the σ/π shape distinction is exactly the one used here, and it is why protein backbones adopt the discrete φ/ψ dihedral conformations mapped by the Ramachandran plot. Finally, magnetic resonance and photochemistry of O2, NO, and related radicals rely on identifying which symmetry-labeled orbital houses the unpaired electron, since that assignment controls both the ground-state term symbol and the selection rules for electronic transitions.

Ask the tutor: Why does s–p mixing push the 3σg orbital above the 1πu level for Li2 through N2, but not for O2 and F2?

Start from the atomic orbitals available on two like atoms lying on the z-axis: 1s, 2s, 2px, 2py, 2pz on each center (and beyond, for heavier atoms). The molecular Hamiltonian for the electrons moving in the field of the two fixed nuclei has a special property that atoms alone do not: it is invariant under any rotation about the internuclear axis, and — for a homonuclear diatomic — it is also invariant under inversion through the bond midpoint, and under reflection through any plane containing the axis. These are the symmetry operations of the point groups C∞v (heteronuclear, e.g. CO, HF) and D∞h (homonuclear, e.g. N2, O2). Because the Hamiltonian commutes with the rotation operator about the bond axis, its eigenfunctions — the molecular orbitals — can be chosen as eigenfunctions of the z-component of orbital angular momentum, with eigenvalue $m_\ell \hbar$ where $m_\ell = 0, \pm1, \pm2, \dots$ The magnitude $|m_\ell|$ is what the Greek-letter labels encode:

• $|m_\ell| = 0$: σ orbital — cylindrically symmetric, no angular dependence around the axis, no nodal plane containing the bond axis.
• $|m_\ell| = 1$: π orbital — one nodal plane containing the bond axis; always comes as a degenerate pair (e.g. built from $p_x$ and from $p_y$, or their $\pm$ combinations).
• $|m_\ell| = 2$: δ orbital — two nodal planes containing the axis; relevant for d-orbital-based bonding in transition-metal complexes, not needed for main-group diatomics.

Superimposed on this is the parity label, which exists only when the molecule has a genuine center of inversion — true for homonuclear diatomics (D∞h) but not for heteronuclear ones (C∞v, no inversion center because the two nuclei differ). Applying the inversion operator $\hat{i}$ (which sends $\mathbf{r} \to -\mathbf{r}$ through the bond midpoint) to an MO either reproduces the orbital unchanged (gerade, German for "even," subscript g) or flips its sign everywhere (ungerade, "odd," subscript u). Finally, within a given σ or π symmetry class there can be more than one MO — a bonding one and an antibonding one, built from the same pair of atomic orbitals with opposite relative phase. These are distinguished by counting nodes between the nuclei: zero such nodes means constructive overlap and bonding character; one node exactly on the perpendicular bisector plane means destructive overlap and antibonding character, marked with a star (σ, π). Putting the three ingredients together gives the complete alphabet used to describe every diatomic MO: shape (σ, π, δ) + parity (g, u) + bonding/antibonding (plain, *). A typical label reads "$1\pi_u$" or "$2\sigma_g^*$" — shape first, parity as a subscript, star (if present) as a superscript, and often a numbering prefix indicating the order of orbitals of that symmetry type by increasing energy.

Contrast that with orbitals built from the perpendicular p orbitals. Here the lobes overlap sideways, above and below the bond axis, and there is necessarily a nodal plane that contains the internuclear axis itself — the defining feature of π symmetry.

Two structural facts fall directly out of these pictures. First, π orbitals always arrive in degenerate pairs (from $p_x$ and $p_y$, or from $d_{xz}$ and $d_{yz}$) because the molecule has no way to distinguish "up-down" from "front-back" around a cylindrically symmetric axis — this is why filling the π orbitals of O2 with its two "extra" electrons, one in each degenerate π* orbital by Hund's rule, produces the famous paramagnetism of molecular oxygen. Second, σ orbitals are always non-degenerate (only one way to point along the axis with $m_\ell = 0$), so each σ level holds exactly its own pair of electrons in isolation.

The angular-momentum quantum number about the bond axis, obtained because the Hamiltonian is invariant under rotation $\hat{R}_z(\phi)$ by any angle φ about that axis, is:

where:

• $\hat{L}_z$ is the projection of orbital angular momentum onto the internuclear (z) axis,
• $m_\ell$ is its eigenvalue in units of $\hbar$,
• σ, π, δ label $|m_\ell| = 0, 1, 2$ respectively, in direct analogy to atomic s, p, d for $m_\ell = 0,\pm1,\pm2$ about a single center.

Reading: this equation is the rigorous justification for why "σ" and "π" are not loose descriptive words but precise quantum numbers — every valid molecular orbital of a linear molecule must be a simultaneous eigenfunction of the electronic Hamiltonian and of $\hat L_z$, and the eigenvalue sorts it unambiguously into a symmetry class. The inversion operation, defined only when a true center of symmetry exists (homonuclear diatomics), acts as:

where:

• $\mathbf{r}$ is measured from the bond midpoint (the inversion center),
• the $+$ eigenvalue defines gerade symmetry, the $-$ eigenvalue defines ungerade symmetry.

Reading: because $\hat i^2 = \hat 1$, the only possible eigenvalues are $\pm 1$, so every MO of a homonuclear diatomic is either purely g or purely u — there is no "mixed parity" option. This is what makes the g/u label exhaustive and mutually exclusive. The general symmetry-adapted linear combination (LCAO) construction, valid for a bonding/antibonding pair built from one atomic orbital $\chi_A$ on atom A and its symmetry partner $\chi_B$ on atom B, is:

where:

• $\psi_+$ is normally the lower-energy (bonding, 0 internuclear nodes) combination when $\chi_A,\chi_B$ are both s-type or both $p_z$-type pointing the same way along the bond,
• $\psi_-$ is the higher-energy (antibonding, 1 internuclear node) combination,
• $S = \int \chi_A^* \chi_B \, d\tau$ is the overlap integral,
• $N_\pm$ is the normalization constant guaranteeing $\int |\psi_\pm|^2 d\tau = 1$.

Reading: whether $\psi_+$ or $\psi_-$ ends up gerade or ungerade depends on the parity of the individual atomic orbitals themselves under inversion — this bookkeeping subtlety is worked out explicitly below, because it is the single most common source of student error (mislabeling 1σu as "1σg" or vice versa).

Example 1 (foundational): classify all four MOs formed from 2s orbitals and axial 2pz orbitals on a homonuclear diatomic. Take two atoms A and B on the z-axis. The atomic orbitals available for σ-symmetry combinations are $2s_A, 2s_B$ (each spherically symmetric, hence automatically $m_\ell=0$) and $2p_{zA}, 2p_{zB}$ (each already aligned along the bond axis, hence also $m_\ell = 0$ — a $p_z$ orbital has no angular dependence around the z-axis itself). Four LCAOs result:

Applying the Step 1–4 parity rule from the first derivation: $\psi_1$ (bonding, s-based) is $2\sigma_g$; $\psi_2$ (antibonding, s-based) is $2\sigma_u^*$; $\psi_3$ (bonding, $p_z$-based — note the sign flip rule from Step 2 of the derivation) is $3\sigma_u$; $\psi_4$ (antibonding, $p_z$-based) is $3\sigma_g^*$. All four are σ, none are π, because every atomic orbital involved has zero angular dependence about the bond axis. Node count confirms energy ordering within each g/u sub-pair independently: $\psi_1$ has zero internuclear nodes (lower energy than $\psi_2$'s one internuclear node), and separately $\psi_3$ has zero internuclear nodes (lower energy than $\psi_4$'s one internuclear node). As a limiting-case check, in the united-atom limit where the internuclear distance $R \to 0$, all four orbitals must smoothly connect to atomic orbitals of a single (fictitious) atom of the combined nuclear charge — $\psi_1 \to$ a 2s-like orbital, $\psi_3 \to$ a $2p$-like orbital lying along the axis — consistent with the fact that σ orbitals, having $m_\ell = 0$, connect only to atomic s, p_z, d_{z^2}, … orbitals of the united atom, never to $p_x,p_y$ or other $m\neq0$ states.

Example 2 (applied): use the labels to rationalize the bond order and magnetism of the superoxide ion O2⁻. Neutral O2 has 16 electrons; using the node-counting-derived filling order appropriate for O2 (where $3\sigma_g$ falls below $1\pi_u$, since O is heavy enough that s–p mixing is weak — this ordering swap is addressed quantitatively elsewhere but is taken as given here):

The last two electrons occupy the doubly-degenerate $1\pi_g^*$ level. By Hund's rule, they occupy the two degenerate π* orbitals singly with parallel spins, giving O2 a $^3\Sigma_g^-$ ground term and its well-known paramagnetism. Bond order is:

wait — check the count carefully: bonding electrons are those in $1\sigma_g, 2\sigma_g, 3\sigma_g, 1\pi_u$: $2+2+2+4 = 10$; antibonding electrons are those in $1\sigma_u^*, 2\sigma_u^*, 1\pi_g^*$: $2+2+2 = 6$. So

matching the textbook O=O double bond. Now add one electron to form superoxide O2⁻ (17 electrons): the extra electron enters the remaining empty slot of the $1\pi_g^*$ pair, giving $(1\pi_g^*)^3$ — three electrons in a doubly-degenerate level, necessarily one unpaired spin (paramagnetic, radical anion). Antibonding electron count rises from 6 to 7:

consistent with the experimentally longer, weaker O–O bond in superoxide (bond length ≈ 1.28 Å versus 1.21 Å in O2) — the symmetry-label bookkeeping directly predicts a measurable bond-length trend. As a limiting-case sanity check, removing electrons instead (forming $O_2^+$, dioxygenyl) should raise the bond order back toward 2.5 by emptying part of $1\pi_g^*$ — and indeed $O_2^+$ has a shorter bond (≈1.12 Å) than neutral O2, confirming the internal consistency of the antibonding assignment.

Example 3 (exam/competition level): predict whether the HOMO of the linear, centrosymmetric azide ion N3⁻ is g or u, and state its σ/π character, using symmetry alone. N3⁻ is linear and centrosymmetric (point group D∞h, like CO2), with 16 valence electrons to place in π and σ frameworks built from the 2p orbitals of three nitrogen atoms (central N contributes both perpendicular p orbitals to two degenerate π-symmetry-adapted linear combinations spanning all three centers; terminal N atoms' in-plane p orbitals and s orbitals build the σ framework along with the N–N σ bonds). Because the molecule is D∞h, every MO must be classifiable as g or u under inversion through the central nitrogen, exactly as in Derivation 1 but now with three centers instead of two. The three-center π system (built from the $2p_x$ orbitals perpendicular to the axis on all three N atoms, and independently from the $2p_y$ set) has three π molecular orbitals per perpendicular direction: one fully bonding (no nodes between adjacent atoms), one non-bonding (a node exactly at the central atom, with a zero coefficient there — familiar from the allyl-anion-like three-center π system), and one fully antibonding (nodes between each adjacent pair). Applying the same inversion argument as Step 3 of Derivation 1, extended to three centers related by the inversion center at the central N: the fully bonding π MO (symmetric combination of the two terminal-atom lobes, in phase with the central atom) transforms as $\pi_u$; the non-bonding π MO, which by symmetry must have equal-and-opposite coefficients on the two terminal atoms with a node at the (inversion-related) center, transforms as $\pi_g$; the fully antibonding combination is again $\pi_u^*$. With 16 valence electrons distributed among σ(N–N) bonding pairs, lone pairs, and the two perpendicular three-center π systems, the highest-filled π level in azide is the non-bonding $\pi_g$ pair (doubly degenerate, from $p_x$ and $p_y$ non-bonding combinations) — this is the HOMO, and it is gerade with π symmetry, i.e., labeled $1\pi_g$ (non-bonding). This symmetry assignment is exactly why azide's HOMO has no amplitude on the central nitrogen and is not stabilized by head-on σ donation there — a prediction confirmed by the shape of azide's frontier orbitals used in cycloaddition and "click" azide–alkyne chemistry, where the terminal nitrogen atoms (not the central one) carry the reactive nucleophilic π density.

The σ/π/g/u framework is a synthesis rather than a single person's invention, but its clearest early champion in molecular quantum mechanics was Robert S. Mulliken, who systematized molecular term symbols and orbital symmetry notation for diatomic and polyatomic molecules through the late 1920s and 1930s, building directly on the group-theoretic classification of atomic and molecular states pioneered alongside contemporaries like Friedrich Hund. Mulliken's molecular orbital correlation diagrams — connecting separated-atom limits to united-atom limits through the σ, π, δ, g, u labeling exactly as used above — became the standard language of the field, earning him the Nobel Prize in Chemistry in 1966 "for his fundamental work concerning chemical bonds and the electronic structure of molecules by the molecular orbital method."