VSEPR Theory Fundamentals

VSEPR is not a historical curiosity — it is the everyday mental shortcut every chemist uses before reaching for a computer. In drug design, the geometry around a nitrogen or phosphorus center (pyramidal vs. planar, SN = 4 vs. SN = 3) determines whether a molecule can hydrogen-bond correctly into a binding pocket, and medicinal chemists routinely reason with VSEPR-level intuition when sketching how a new scaffold will sit in an active site before any computational docking is run. In materials chemistry, predicting whether a silicate, phosphate, or borate unit is tetrahedral or trigonal planar (SiO₄⁴⁻ vs. BO₃³⁻) — a direct SN calculation — is the starting point for understanding network-forming glasses, zeolite pore geometry, and cement chemistry. In catalysis and organometallic chemistry, VSEPR-style domain counting (extended to include d-orbital and ligand-field effects) helps rationalize why five-coordinate metal complexes prefer trigonal bipyramidal versus square pyramidal geometries, feeding directly into predicting reaction pathways like Berry pseudorotation. In atmospheric and environmental chemistry, the bent geometry of ozone (O₃, SN = 3 with one lone pair on the central oxygen, exactly analogous to SO₂ in the worked examples above) explains its dipole moment and its characteristic UV absorption that protects the stratosphere. In spectroscopy and instrumentation, VSEPR-predicted point group symmetry is the first step in predicting which vibrational modes are IR- or Raman-active, since symmetry-adapted selection rules depend entirely on molecular shape. Even in the classroom of first-year general chemistry, VSEPR remains the single most commonly tested qualitative-reasoning tool in inorganic and organic chemistry alike, precisely because it needs no software, no basis set, and no computer — only a correct Lewis structure and a steric-number count.

Ask the tutor: How do the lone-pair and multiple-bond repulsion strengths get ranked, and why do lone pairs specifically prefer equatorial positions in a trigonal bipyramid?

Start from a completed Lewis structure — this is non-negotiable, because VSEPR is not a rival to Lewis theory but a geometric extension of it. Once you have decided how many bonds and lone pairs sit on a central atom, VSEPR treats every one of those regions of electron density as a domain: a roughly localized, negatively charged cloud that repels every other domain via electrostatics reinforced by the Pauli exclusion principle (electrons of like spin, and generally like charge clouds, avoid close approach). The central atom's nucleus and its core electrons sit at the origin; the domains are the "occupants" trying to get as far from one another as possible while staying attached (for bonding pairs) or hovering close to the nucleus (for lone pairs). Because all domains lie at roughly the same radius from the central atom, the problem is geometrically identical to placing N points on the surface of a sphere so that they are as spread out as possible — a classic point-repulsion (Thomson-type) problem. For small N the solutions are the textbook Platonic-ish arrangements: 2 points go to the poles (180° apart, linear), 3 points go to an equilateral triangle in a plane (120° apart, trigonal planar), 4 points go to the corners of a tetrahedron (109.5° apart), 5 points go to a trigonal bipyramid (a geometry with two distinct angle sets, 90° and 120°, because five points cannot be made perfectly equivalent on a sphere), and 6 points go to an octahedron (90° between adjacent points, 180° between opposite ones). VSEPR theory simply asserts that real electron domains adopt these same idealized arrangements, because the underlying physics (repulsion between charge clouds, extremized subject to a fixed number of clouds) is the same problem. The steric number (SN) is the head-count that tells you which of these point-arrangements applies. It is defined purely from the Lewis structure of the central atom:

Two crucial conventions make this count work correctly. First, a double or triple bond is still only one domain, because all the electron density of a multiple bond — the σ bond plus the accompanying π bond(s) — is confined to the same one direction in space (along the internuclear axis); it does not create a second repelling region pointing elsewhere. Second, a lone pair counts exactly like a bonded atom for the purpose of finding the underlying point-arrangement, even though it will later turn out to repel more strongly than a bonding pair (that refinement is the subject of Lone Pair and Multiple Bond Effects). At the SN stage, all domains are treated as equivalent generic "arms" of the point-repulsion problem. Once SN is known, the electron-domain geometry is read off a short list, and the molecular geometry is obtained by mentally erasing the lone pairs and looking only at where the atoms ended up. When there are no lone pairs, the two geometries coincide (as in CH₄, SN = 4, tetrahedral atoms, tetrahedral electron domains). When lone pairs are present, the molecular geometry is a "sub-shape" carved out of the full point-arrangement — this is exactly why H₂O (SN = 4, tetrahedral electron-domain geometry, two lone pairs) is described as "bent," even though its electron domains are tetrahedral, not bent at all.

The compression of water's bond angle from the idealized 109.5° to the observed 104.5° is a preview of the lone-pair refinement: lone pairs, held by only one nucleus rather than shared between two, are on average closer to the central atom and more angularly diffuse, so they repel more strongly than bonding pairs and squeeze the bonding pairs together. That ordering of repulsion strengths (lone pair–lone pair > lone pair–bond pair > bond pair–bond pair) is developed fully in Lone Pair and Multiple Bond Effects; here the essential point is only that SN = 4 fixes the tetrahedral framework, and the lone pairs then distort the observed angle within that framework.

The entire predictive engine of VSEPR compresses into two statements: a counting rule for SN, and a repulsion-minimization principle that maps SN onto geometry.

where: $n_{\sigma\text{-bonded atoms}}$ is the number of atoms directly σ-bonded to the central atom (each multiple bond still counts as one atom/one domain), and $n_{\text{lone pairs}}$ is the number of non-bonding electron pairs residing on the central atom, both read directly from a correct Lewis structure. Reading: SN is a domain count, not an electron count — it tells you how many repelling regions of space surround the central atom, which is exactly the number of "arms" in the point-repulsion problem that fixes geometry.

where: $q_i, q_j$ are the (negative) effective charges of domains $i$ and $j$, and $r_{ij}$ is the angular/spatial separation between domain $i$ and domain $j$ on the sphere of radius equal to the bonding/lone-pair distance from the nucleus. Reading: the observed geometry is the one that minimizes the total pairwise repulsion energy summed over all domain pairs; since all domains at a given SN are (to leading order) equally charged, the minimum-energy arrangement is the most symmetric one — the same arrangement solved by the classical problem of spreading N charges over a sphere.

where: $\theta_{\text{ideal}}$ is the tetrahedral angle obtained by placing four identical point charges as far apart as possible on a sphere — the unique symmetric solution for four mutually repelling points. Reading: this is the single most quoted number in VSEPR, and it is not an empirical fit — it falls straight out of the geometry of a regular tetrahedron, as the next derivation shows explicitly.

Example 1 (foundational): Ammonia, NH₃. Nitrogen has 5 valence electrons; three hydrogens contribute 1 each, total $5 + 3 = 8$ electrons = 4 pairs. Three N–H σ bonds use 3 pairs; the remaining pair is a lone pair on nitrogen. Steric number:

Electron-domain geometry: tetrahedral (four domains, idealized 109.5°). Molecular geometry: trigonal pyramidal (three atoms + one lone pair visible as the "missing" vertex). The experimental H–N–H angle is 107.3°, compressed from the ideal 109.5° by the lone pair, exactly the same qualitative effect seen in water but weaker, because NH₃ has only one "extra-repulsive" lone pair compared to water's two. As a limiting-case check: if the lone pair were (hypothetically) replaced by a fourth bonding pair to an atom of similar electronegativity, the molecule would relax to a perfectly symmetric tetrahedron (109.5° exactly), which is precisely what happens on going from NH₃ to NH₄⁺ — protonating the lone pair converts it into a fourth N–H bond, SN stays at 4, but now all four domains are bonding pairs, and NH₄⁺ is experimentally a perfect tetrahedron with all H–N–H angles equal to 109.5°.

Example 2 (applied): Predicting the shape of the sulfate ion, SO₄²⁻, and comparing to SO₂. For SO₄²⁻: sulfur contributes 6 electrons, four oxygens contribute $4\times 6 = 24$, plus 2 for the $2-$ charge, total $6+24+2 = 32$ electrons = 16 pairs. A Lewis structure consistent with formal charges close to zero places two S=O double bonds and two S–O single bonds (with the singly-bonded oxygens each carrying a $-1$ formal charge, and resonance distributing the double-bond character equally over all four oxygens in the real ion). Regardless of exactly how the 16 pairs are formally partitioned among the resonance structures, sulfur is σ-bonded to four oxygen atoms and has zero lone pairs on itself:

SO₄²⁻ is therefore a perfect (electron-domain = molecular) tetrahedron, and because all four S–O bonds are rendered equivalent by resonance/symmetry, the observed O–S–O angles are all 109.5°, and all four S–O bond lengths are equal (~149 pm, intermediate between typical S–O single and S=O double bond lengths) — a beautiful confirmation that VSEPR geometry is controlled by domain count, not by which specific resonance structure you happened to draw. Contrast this with SO₂: sulfur contributes 6, two oxygens contribute 12, total 18 electrons = 9 pairs. A valid Lewis structure has one S=O double bond, one S–O single bond (via a resonance-averaged picture, both S–O bonds are actually equivalent, each with partial double-bond character), and one lone pair on sulfur to complete its octet (formal charges: 0 on S if drawn with one double and one dative/single bond in the standard resonance treatment). Sulfur's steric number:

Three domains give a trigonal planar electron-domain geometry, and with one domain being a lone pair, the molecular geometry is bent, with an experimental O–S–O angle of 119.5° — very close to the ideal 120° because there is only a single, moderately repulsive lone pair doing the compressing (compare this modest 0.5° compression to water's 5° compression from its four-domain, two-lone-pair tetrahedral parent). This pairing of examples shows the double role of SN: SO₄²⁻ (SN=4, no lone pairs) is a real tetrahedron, while SO₂ (SN=3, one lone pair) is bent — despite both being sulfur oxoanions/oxides, their steric numbers, not their formulas, control the shape.

Example 3 (exam/competition level): Determining the geometry of the triiodide ion, I₃⁻, and predicting the effect of substituting one terminal iodine by a less symmetric ligand. Iodine contributes 7 valence electrons; three iodine atoms contribute $3 \times 7 = 21$, plus 1 electron for the $1-$ charge, total $22$ electrons = 11 pairs. In I₃⁻ the central iodine is bonded to the two terminal iodines by single bonds (2 pairs, 4 electrons used), and the remaining $22 - 4 = 18$ electrons (9 pairs) are distributed as three lone pairs on each terminal iodine ($2\times 3=6$ pairs) plus $9-6=3$ lone pairs on the central iodine. The central iodine's domain count:

Five domains give a trigonal bipyramidal electron-domain geometry. The three lone pairs, being the most repulsive domains, preferentially occupy the three equatorial sites (each equatorial site has only two 90° neighbors — the two axial bonds — whereas an axial lone pair would suffer three 90° contacts from the equatorial domains); this leaves the two bonded iodine atoms in the two axial positions, 180° apart. The molecular geometry is therefore linear, and indeed I₃⁻ is experimentally linear (I–I–I = 180°), a textbook result quoted in every discussion of hypervalent halogen species and central to explaining why starch–iodine complexes and polyiodide chemistry favor linear chains. As a check on the reasoning, consider the asymmetric hypothetical species ICl₂⁻ built the same way: central I bonded to two Cl atoms, with the same domain bookkeeping — iodine contributes 7, two chlorines contribute 14, plus 1 for the charge, giving 22 electrons total, identical electron count to I₃⁻ once the different atomic identities are accounted for in the same "2 bonded atoms + 3 lone pairs on I" pattern. The central iodine again reaches SN = 5 with 3 lone pairs equatorial, and the molecular geometry is again linear (Cl–I–Cl = 180°), confirming that VSEPR's prediction depends on the steric number of the central atom, essentially independent of whether the terminal atoms are identical (I₃⁻) or different (ICl₂⁻) — the terminal atoms' identity affects bond lengths and polarity, but not, to leading order, the domain-count-driven skeleton.

VSEPR as a systematic predictive framework is usually credited jointly to Ronald Gillespie and Ronald Nyholm, who in the 1950s-60s took earlier, more scattered observations (notably by Nevil Sidgwick and Herbert Powell in the 1940s, who first noted that lone pairs occupy space and influence geometry) and turned them into a coherent, quantitatively testable set of rules. Gillespie, working at University College London and later McMaster University in Canada, spent decades refining the ordering of repulsion strengths (lone pair–lone pair > lone pair–bond pair > bond pair–bond pair) and testing it exhaustively against microwave spectroscopy and electron/X-ray diffraction bond-angle data across the periodic table. Nyholm, his collaborator and a transition-metal spectroscopist at UCL, contributed the broader inorganic perspective and pushed the theory's application beyond simple main-group hydrides toward genuinely diverse coordination environments. What makes the Gillespie–Nyholm synthesis (published as the "Nyholm–Gillespie theory" in the mid-1950s and popularized through Gillespie's subsequent decades of textbook writing) historically important is its astonishing predictive economy: with nothing more than a Lewis structure and a head-count, it correctly anticipates bond angles across hundreds of molecules to within a few degrees, using no quantum mechanical calculation whatsoever. Gillespie himself was always careful to describe VSEPR not as a rigorous derivation from the Schrödinger equation but as an extremely successful model — a simplified, physically motivated bookkeeping scheme whose success he attributed to the dominance of Pauli-exclusion-driven electron correlation between same-spin, same-shell electron pairs, a point later given more rigorous quantum-chemical grounding by others using the Laplacian of the electron density (the basis of the related Electron Localization Function / AIM analyses).