Molecular Geometry & VSEPR · Guide · glidepathscience.org
GuideMolecular Geometry & VSEPR
Contents
This Guide explains the chemistry behind every Walkthrough step, Concept Challenge question and Predict round, in the order a textbook would take it. Part 1 prepares the Intro track and Part 2 the Advanced track; each takes about fifteen minutes to read. Each section has a worked example, chosen outside the Walkthrough steps and the Predict rounds.
Part 1 · The essentials
Valence electrons and Lewis structures
Carbon dioxide drifts out of a soda as a gas, while water, a lighter molecule, is a liquid. Much of the difference is shape: CO₂ is straight and H₂O is bent, so water molecules are polar and hold one another by hydrogen bonds, while carbon dioxide's do not. Solubility, boiling points and the way enzymes recognize molecules all depend on shape, and predicting it starts with a count of electrons.
Only an atom's outermost electrons take part in bonding. These valence electrons are the s and p electrons of the outer shell, the one with the highest principal quantum number n. Sulfur, [Ne] 3s² 3p⁴, has six, and its Lewis symbol shows them as six dots around the letter S. The count is the group number in groups 1 and 2, and the group number minus ten in groups 13 to 18. A molecule or ion has the sum of its atoms' valence electrons, adjusted for its charge:
$\text{valence electrons} = \sum \text{outer-shell electrons} - \text{charge}$
Subtracting a negative charge adds electrons: an anion has extra ones. A Lewis structure then shares these electrons out. Put the least electronegative atom other than hydrogen in the center, since hydrogen forms only one bond. Join each outer atom to the center with a single bond, a shared pair drawn as a line. Complete each outer atom's octet with lone pairs (hydrogen needs only its one pair). Whatever electrons remain go on the central atom as lone pairs. If the center is still short of an octet, turn lone pairs on outer atoms into double or triple bonds (boron and beryllium are often left with fewer than eight, as in BF₃). Atoms from the third period down may hold more than eight electrons, an expanded octet; second-period atoms never do, because their valence shell has room for only eight.
The two usual slips are counting every electron of an atom instead of its outer shell (sixteen for sulfur instead of six) and getting the charge's sign backward. A correct count still leaves choices about where the electrons go, and the next section is about choosing.
Formal charge and resonance
Often more than one arrangement of bonds gives every atom an octet. To decide which drawing describes the species best, chemists use a piece of bookkeeping called formal charge. Pretend that every bond is shared exactly evenly: each atom owns all of its lone-pair electrons and half of each bonding pair. Compare that with the atom's own valence electrons:
$\text{formal charge} = \text{valence electrons} - \text{lone-pair electrons} - \text{bonds}$
Here a double bond counts as two bonds. The formal charges in a structure always add up to the charge of the species. Among valid structures, prefer the one whose formal charges are closest to zero, and the one that puts any negative formal charge on the more electronegative atom.
Sometimes two or more structures are equally good and differ only in where a multiple bond sits. These are resonance structures. The name misleads: the species does not flip between them. It has one structure, in which the bonds the drawings disagree about are identical, with a bond order equal to the average. In the formate ion, HCO₂⁻, the double bond can be drawn to either O, so each C–O bond has order 3/2 and both are the same length.
Formal charges are bookkeeping, not the real partial charges on the atoms. For shape, the example's last sentence matters most: every reasonable structure gives the central atom the same number of electron regions, and that number sets the shape.
Counting electron domains
This sim's theory of shape is the valence-shell electron-pair repulsion model, VSEPR. It treats the electrons around a central atom as a few regions of concentrated negative charge, electron domains, that repel one another and settle as far apart as they can.
Each atom bonded to the center is one domain, whatever the bond order: the two or three pairs of a multiple bond all lie between the same two atoms, so they point the same way. Each lone pair on the central atom is one more domain, and so is a single unpaired electron in a radical. Lone pairs on the outer atoms do not count; they belong to those atoms.
$\text{electron domains} = \text{bonded atoms} + \text{lone pairs}$
Chemists summarize a central atom as AXₘEₙ: A is the central atom, m the number of bonded atoms X and n the number of lone pairs E, so the oxygen of HOF is AX₂E₂. When the central atom's formal charge is zero, a quick way to find n is to take the central atom's valence electrons, subtract one for each bond (two for a double bond, three for a triple), and halve what is left.
The usual mistakes are counting electrons instead of regions (eight instead of four for the oxygen of HOF), counting a double bond as two domains, and forgetting the lone pairs. With the count in hand, the next question is how the domains arrange themselves in space.
Pairs spread out in three dimensions
A Lewis structure is flat: methane is drawn as a cross with its H atoms 90° apart. In three dimensions four domains can get much farther apart, and they do.
Picture the domains as points free to slide over a sphere around the central atom, each pushing the others away. Two points settle on opposite sides, 180° apart: a linear arrangement. Three spread around an equator at 120°: trigonal planar. Four go to the corners of a regular tetrahedron, 109.5° apart. The tetrahedral angle comes from symmetry alone. Four equal unit vectors pointing from the center to the corners add to zero, so the dot product of one of them with the sum is zero too: $1 + 3\cos\theta = 0$, which gives $\cos\theta = -1/3$ and θ = 109.47°. Three vectors in a plane give $1 + 2\cos\theta = 0$, or 120°.
Five domains make a trigonal bipyramid, which has two kinds of position. Two axial positions sit on opposite poles, and three equatorial positions form a triangle around the middle. Equatorial domains are 120° from each other and 90° from the axial ones. Six domains make an octahedron, every domain 90° from four others and 180° from the sixth. These five arrangements are the electron-pair geometries.
The classic error is to picture four domains as a flat square, as the Lewis drawing suggests; the other is to treat the five positions of a trigonal bipyramid as equivalent. That difference between positions matters as soon as some of the domains are lone pairs.
Molecular shape: the atoms alone
Experiments that find molecular structures locate the nuclei, so the name of a shape describes where the atoms are. Chemists keep the two ideas apart: the electron-pair geometry counts every domain, and the molecular structure, or shape, names the arrangement of the atoms alone. When there are no lone pairs the two names are the same. Ammonia has four domains, so its electron pairs are tetrahedral, but with one corner taken by a lone pair the atoms form a trigonal pyramid, N at the apex. Water, with two lone pairs among its four domains, is bent.
With five domains the lone pairs have a choice of positions, and they take the equatorial ones first. The strongest repulsions are between domains 90° apart. An axial position has three neighbors at 90°; an equatorial position has only two. Lone pairs repel more strongly than bonding pairs, so the arrangement with the fewest lone-pair contacts at 90° wins. One lone pair (AX₄E) leaves a seesaw, two (AX₃E₂) a T shape, and three (AX₂E₃) a linear molecule, with the outer atoms on the axis. With six domains, one lone pair (AX₅E) leaves a square pyramid; a second goes opposite the first, 180° away rather than 90°, leaving a square planar AX₄E₂ molecule.
The usual mistake is to name the electron-pair geometry when the shape is asked for, calling ammonia tetrahedral; the other is to put a lone pair in an axial position. Lone pairs do more than choose their positions, though: they also bend the bonds around them.
Lone pairs and multiple bonds squeeze the angles
Methane, ammonia and water all have four electron domains, but their measured angles are 109.5°, 106.7° and 104.5°. The ideal geometries give the starting angles; the kinds of domain decide how far they move.
A bonding pair is held between two nuclei and stretched out along the bond. A lone pair is held by only one nucleus, so it sits closer to the central atom and spreads over a wider angle. It therefore pushes harder on its neighbors than a bonding pair does. The repulsions fall in the order lone pair–lone pair, then lone pair–bonding pair, then bonding pair–bonding pair, and each lone pair closes the angles between the bonds a little more. A double or triple bond carries more electron density than a single bond, so its domain is larger too: in formaldehyde, H₂CO, the H–C–H angle closes to 116.5°, below the trigonal 120°. In the other direction, a single unpaired electron is a smaller domain than a pair, and very large outer atoms can crowd each other; either can open an angle past its ideal value.
A common slip is to reason that lone pairs push the bonds apart and open the angle; they push the bonds away from themselves, which crowds them together. With shape and angles known, the next question is how the charge is spread.
Polar bonds and polar molecules
Polarity decides whether a substance mixes with water, how high it boils and how it lines up in an electric field. It begins with how evenly a bond's electrons are shared.
Electronegativity, χ, measures how strongly an atom in a molecule attracts the electrons it shares. On Pauling's scale fluorine has the highest value, 3.98; oxygen has 3.44, chlorine 3.16, nitrogen 3.04, carbon 2.55 and hydrogen 2.20. The difference between the two atoms of a bond measures how polar the bond is:
$\Delta\chi = \chi(\text{more electronegative}) - \chi(\text{less})$
The shared electrons spend more time near the more electronegative atom, which takes a partial negative charge, δ−, and leaves δ+ on the other. Chemists draw a bond dipole as an arrow along the bond, crossed at its tail, pointing toward the δ− atom, with its length growing with Δχ. Physicists draw the dipole moment the other way, from − to +; this sim uses the chemists' convention.
A molecule is polar only if its bond dipoles, added as vectors tip to tail, leave a net dipole. In symmetric shapes they cancel: linear AX₂, trigonal planar AX₃, tetrahedral AX₄, trigonal bipyramidal AX₅, octahedral AX₆, and also square planar AX₄E₂ and linear AX₂E₃, as long as all the X atoms are the same. Replace one X with a different atom, or bend the molecule, and the cancellation fails.
Two mistakes are common: concluding that a molecule with polar bonds must be polar, and concluding that a molecule with lone pairs must be polar (XeF₄ has two, and is nonpolar). Deciding whether arrows cancel means seeing them in three dimensions, which raises the question of how to draw three dimensions on paper.
Drawing in three dimensions: wedges and dashes
Paper is flat and molecules are not, so chemists draw depth with a code. A plain line is a bond in the plane of the page. A solid wedge, widening toward the outer atom, is a bond coming toward you. A hashed wedge is a bond pointing away from you, behind the page.
A tetrahedral center is usually drawn with two plain lines, one wedge and one hashed wedge: two bonds lie in the page, and the plane of the other two stands perpendicular to it, one bond in front and one behind. The drawing records one point of view: turn the molecule and another pair of bonds lies in the page, though the molecule has not changed.
That last point is the usual mistake: an angle measured on a wedge–dash drawing is not the angle in the molecule, unless both bonds lie in the page. With Part 1's tools in hand, Part 2 asks why the rules work and where they stop working.
Part 2 · Going further
Points on a sphere
Strip VSEPR to its core and it becomes a problem in geometry: N equal points that repel one another and can move only over a sphere. Where do they come to rest? With a 1/r repulsion this is J. J. Thomson's problem from his early model of the atom.
Give each domain i a direction $\hat{u}_i$ from the central atom, a distance from it $\rho_i$ (in bond lengths) and a weight $w_i$, and let every pair repel with an inverse power n of their distance. The energy is
$U = \sum w_i w_j / |\rho_i \hat{u}_i - \rho_j \hat{u}_j|^{n}$
summed over every pair once. For equal points, all ρ and w are 1. For two, three, four and six points the lowest arrangements are the electron-pair geometries for any reasonable law. Five points are the delicate case. The trigonal bipyramid is lowest only for repulsions softer than about $1/r^{15}$, and the best square pyramid is never far above it: 0.37 % with $1/r^2$ and at most 1.7 % for any law up to that crossover. A small difference means an easy path from one to the other. In PF₅ a twist through a square pyramid, the Berry pseudorotation, swaps axial and equatorial F atoms so fast that its ¹⁹F NMR spectrum shows one kind of F. Seven points form a pentagonal bipyramid for $1/r$ and $1/r^2$; for stiffer laws, such as $1/r^{2.5}$ and $1/r^3$, a less symmetric arrangement wins, but only by 0.0013 % and 0.009 %. Seven-domain molecules such as IF₇ are correspondingly floppy.
A small energy gap is not a sign of rigidity; it means the opposite. Real molecules, though, are not made of equal points.
A lone pair in the model, and where the model misses
To model molecules, lone pairs and multiple bonds must take more room than single bonds. How to give it is a choice, a hypothesis to test against measured molecules.
This sim puts each bonding domain at distance 1 from the center and each lone pair closer in, at $\rho = 0.70$ (nearer the nucleus, it spreads wider), with the repulsion law $n = 2$. A double bond gets weight 1.2, a triple bond 1.4 and a resonance-averaged order of 3/2 weight 1.1; a single electron sits at distance 1 with weight 0.5. The lone-pair distance and the law were fitted to ammonia and water together: the model gives 106.6° and 105.1° against the measured 106.7° and 104.5°. With $1/r^2$ no choice of ρ reaches 104.5° for water: the angle falls to 104.9° near ρ = 0.65, then opens again, because a lone pair very near the center pushes on every bond alike. The simpler idea, lone pairs at the bond distance that repel harder (weight 1.4), comes close as well, at 105.8° and 102.8°, but it puts the two lone pairs of XeF₄ side by side, 96° apart, where the real XeF₄ is square planar.
Even the better model has limits. In SF₄ it places the lone pair equatorial, as measured, and its F(ax)–S–F(ax) angle of 171.3° is close to the measured 173.1°. But it opens F(eq)–S–F(eq) to 121.2°, just past the ideal 120°, while the real angle closes to 101.6°. The qualitative rule that lone pairs squeeze their neighbors gets that direction right; equal repulsions between points do not.
The mistake to avoid is trusting a model because it matches the molecules it was fitted to. The test is how it does on the others, and the misses point to what it leaves out: the sizes and electronegativities of the atoms.
Hybrid orbitals: Coulson's angles and Bent's rule
Down group 16 the hydrides' angles drop from 104.5° for H₂O to 92.1° for H₂S and 90.3° for H₂Te, and down group 15 from 106.7° for NH₃ to 93.3° for PH₃. VSEPR says only “less than 109.5°”. Valence bond theory says more, by describing each bond as an overlap between an outer atom and a hybrid orbital on the central atom, a mixture of its s and p orbitals.
A hybrid written spⁿ has one part s to n parts p, so its s character is 1/(1 + n). Coulson showed that two equivalent spⁿ hybrids point at an angle θ fixed by n (this n is the hybrid's p-to-s ratio, not the repulsion power of the model):
$\cos\theta = -1/n$ $\text{s character} = 1/(1+n)$
So sp hybrids are 180° apart, sp² 120° and sp³ 109.47°, and pure p orbitals (n very large) meet at 90°. In the heavier elements the valence s orbital is much lower in energy than the p orbitals and different in size, so mixing them pays off less. The bonds use nearly pure p orbitals, near 90°, and the lone pairs keep the s character. Bent's rule says where the s character goes within one molecule: toward electropositive groups and lone pairs, while bonds to electronegative atoms get more p character and close up. NF₃'s F–N–F angle (102.2°) is smaller than NH₃'s H–N–H (106.7°), and in CH₂F₂ the F–C–F angle closes to 108.3° while H–C–H opens to 113.7°. The rule is a tendency: PF₃ (97.8°) is wider than PH₃ (93.3°).
Two slips are common here: dropping the minus sign (cos θ is negative above 90°, and n must come out positive), and using 1/n where the s character is 1/(1 + n). The success of hybrids raises a fair question: do they cause these shapes?
Hybrids describe; they do not cause
It is tempting to say that carbon in methane is sp³ hybridized and therefore tetrahedral. The logic runs the other way. A molecule's shape is the arrangement of lowest total energy. Hybrid orbitals are one convenient way of dividing its electron density into pieces that match that shape, and there are others. Water's two lone pairs can be described as two equivalent “rabbit ears”, or as two different orbitals: one s-rich in the plane of the molecule and one pure p perpendicular to it. The second description matches water's photoelectron spectrum, in which its two highest-energy pairs leave at different energies, 12.6 eV and 14.7 eV; both descriptions add up to the same density.
The same care applies to molecules whose Lewis structures place more than eight electrons around the center, such as SF₆ and XeF₂. Older texts explained these expanded octets with sp³d² and sp³d hybrids that use d orbitals. Modern calculations give the d orbitals only a small polarizing role and describe the bonding instead with charge separation and three-center four-electron bonds: in XeF₂, one bonding pair spread over all three atoms, F–Xe–F, plus a nonbonding pair on the two F atoms. VSEPR is untouched by this debate. It counts regions of electron density around the center, and that count is the same in every description.
The usual error is to treat a hybrid label as a cause (“sp³ forces 109.5°”) or as an electron count (“sulfur has twelve valence electrons in SF₆”; it has six). Polarity has a similar subtlety: the shape tells whether a dipole exists, but not how large it is.
How big is a dipole?
A charge +q and a charge −q a distance d apart have a dipole moment μ = qd. Molecular dipoles are measured in debye, D:
$1\unit{D} = 3.336\times 10^{-30}\unit{C·m}$ $e = 1.602\times 10^{-19}\unit{C}$
For a bent AX₂ molecule with two equal bond dipoles $\mu_b$ at an angle θ, the components across the molecule's axis cancel and those along it add. Each bond makes θ/2 with the axis, so
$\mu = 2\mu_{b}\cos(\theta/2)$
Running this backward from a measured μ splits the molecule's dipole into bond dipoles, but the result is bookkeeping, because lone pairs carry dipoles of their own. Ammonia and nitrogen trifluoride show it. Both are pyramidal, and the N–F bonds are more polar than N–H, yet NH₃'s dipole is 1.47 D and NF₃'s only 0.23 D. The N–H dipoles point toward N, the same way as the lone pair's, and add to it; the N–F dipoles point away from N and nearly cancel it. For a diatomic molecule, comparing μ with the dipole of a full charge at the bond length gives a rough percent ionic character:
$\text{ionic character} = \mu/(e\,d)$
The two common slips are losing the factor 2 and using θ where the formula needs θ/2. Some molecules break more than the size of a dipole: they break VSEPR itself.
Where VSEPR fails, and molecular orbitals
VSEPR is a model of covalent molecules of main-group elements, and it fails in four main ways. In mostly ionic molecules the electrons around the anion are nearly spherical, with no directed pairs: gaseous Li₂O is linear although its O has two lone pairs, while CaF₂, SrF₂ and BaF₂, with no lone pairs on the metal, are bent, because the F⁻ ions distort the large metal ion's own inner electrons, an effect a count of valence pairs cannot see. In some heavy-atom compounds the lone pair stays in a spherical s orbital and takes no direction: [TeCl₆]²⁻ and [BrF₆]⁻ are regular octahedra, and gaseous SnCl₂ is bent near 95–99°, not just under 120°, because its two bonds use nearly pure p orbitals, as in H₂Te. Unusual bonding can remove a lone pair from play: N(SiH₃)₃ is planar where N(CH₃)₃ is pyramidal. And transition-metal complexes follow their d orbitals and d electrons: [Ni(CN)₄]²⁻ is square planar, and W(CH₃)₆, with no d electrons at all, is a trigonal prism.
Molecular orbital theory builds shapes from orbitals spread over the whole molecule. A Walsh diagram follows each orbital's energy as an $\mathrm{AH}_2$ molecule bends. Bending lowers one orbital that is empty when the molecule has four valence electrons or fewer and occupied when it has more, which gives Walsh's rule: $\mathrm{AH}_{2}$ is linear with up to 4 valence electrons and bent with 5 to 8. The same theory explains what no Lewis structure can, such as the two unpaired electrons that make O₂ paramagnetic.
The mistake to avoid is treating VSEPR as a law rather than a model: it predicts the shapes of most molecules a chemistry course meets, and its failures are where the more complete theories begin.
Reference
Common mistakes
- Counting all the electrons, or the charge backward. Only the outer shell counts (S has 6), and a negative ion has extra electrons.
- Counting a double bond twice. Every bonded atom is one domain, whatever the bond order.
- Forgetting the lone pairs. They are domains, though no atom sits there.
- Naming the electron pairs instead of the atoms. NH₃'s pairs are tetrahedral; its shape is a trigonal pyramid.
- Putting a lone pair axial. In a trigonal bipyramid lone pairs go equatorial first.
- Letting lone pairs open the angle. They squeeze the bonds together.
- Polar bonds, so a polar molecule. Symmetric shapes cancel their bond dipoles.
- Using 1/n for the s character. It is 1/(1 + n), with n from $\cos\theta = -1/n$.
- Losing the 2 or the half-angle. $\mu = 2\mu_b\cos(\theta/2)$.
Equation sheet
| Part 1 | |
|---|---|
| valence | $\text{valence electrons} = \sum \text{outer-shell electrons} - \text{charge}$ |
| formal charge | $\text{formal charge} = \text{valence electrons} - \text{lone-pair electrons} - \text{bonds}$ |
| domains | $\text{electron domains} = \text{bonded atoms} + \text{lone pairs}$ (a multiple bond counts once) |
| geometries | 2 linear 180° · 3 trigonal planar 120° · 4 tetrahedral 109.5° · 5 trigonal bipyramidal 90°/120° · 6 octahedral 90° |
| polarity | $\Delta\chi = \chi(\text{more electronegative}) - \chi(\text{less})$ |
| Part 2 | |
|---|---|
| model | $U = \sum w_i w_j / |\rho_i \hat{u}_i - \rho_j \hat{u}_j|^{n}$, lone pairs at $\rho = 0.70$, $n = 2$ |
| Coulson | $\cos\theta = -1/n$ for two equivalent $\mathrm{sp}^{n}$ hybrids; $\text{s character} = 1/(1+n)$ |
| dipoles | $\mu = 2\mu_{b}\cos(\theta/2)$; $\text{ionic character} = \mu/(e\,d)$, $d$ the bond length |
| units | $1\unit{D} = 3.336\times 10^{-30}\unit{C·m}$, $e = 1.602\times 10^{-19}\unit{C}$ |
| Walsh | $\mathrm{AH}_{2}$: up to 4 valence electrons linear; 5 to 8 bent |
Symbols: AXₘEₙ central atom A with m bonded atoms X and n lone pairs E · χ electronegativity (Pauling) · δ+, δ− partial charges · θ bond angle · n the p-to-s ratio of an spⁿ hybrid, or the power of the repulsion law · $\hat{u}_i$, $\rho_i$, $w_i$ a domain's direction, distance (in bond lengths) and weight · μ dipole moment, $\mu_b$ one bond's · e the elementary charge · d bond length. 1 pm = 10⁻¹² m.
Further reading
OpenStax, Chemistry 2e, §7.2–7.4 (covalent bonding and electronegativity, Lewis structures, formal charges and resonance) and §7.6 (molecular structure and polarity), whose order and terms Part 1 follows, and §8.2 (hybrid atomic orbitals), for Part 2's section on hybrids; R. J. Gillespie and I. Hargittai, The VSEPR Model of Molecular Geometry, for the model and its limits. The Field Manual lists every source of the measured values.
Molecular Geometry & VSEPR v1.4 · © 2026 Dr. David Meier · CC BY-NC-SA 4.0 · glidepathscience.org