Geometria Molecular

Predicting molecular shapes isn't as simple as counting electrons

I spent years teaching introductory chemistry, and molecular geometry still trips people up because most textbooks present it like a formula you plug numbers into. It's not. The theory works for simple main-group compounds, but the moment you hit transition metals or heavy p-block elements, the rules start breaking. What geometria molecular actually measures You're describing the three-dimensional arrangement of atoms around a central atom. The standard framework is VSEPR—Valence Shell Electron Pair Repulsion—which assumes electron domains repel each other and spread out as far as possible. Two domains give you linear geometry at 180 degrees. Three domains give trigonal planar at 120 degrees. Four domains give tetrahedral at 109.5 degrees. Five and six domains follow the same logic but produce trigonal bipyramidal and octahedral arrangements respectively.

The catch is that VSEPR treats bonding pairs and lone pairs as separate categories with different repulsion strengths. Lone pair-lone pair repulsion is strongest, followed by lone pair-bonding pair, followed by bonding pair-bonding pair. This is why water bends to about 104.5 degrees instead of staying at the perfect tetrahedral angle. The two lone pairs push the hydrogens closer together.

Counting electron domains correctly

Here's where most students lose marks: you don't count individual bonds. You count electron domains. A double bond counts as one domain. A triple bond counts as one domain. A lone pair counts as one domain. That's it. For methane, CH4, you have four bonding domains and zero lone pairs. Tetrahedral geometry. For ammonia, NH3, you have three bonding domains and one lone pair. The electron geometry is still tetrahedral, but the molecular geometry is trigonal pyramidal because you only look at where the atoms sit, not where the lone pair sits. For water, H2O, you have two bonding domains and two lone pairs. Electron geometry is tetrahedral. Molecular geometry is bent or angular. The difference between electron geometry and molecular geometry is the single most important distinction in this topic. Electron geometry includes lone pairs. Molecular geometry describes only the atom positions. They match when there are zero lone pairs on the central atom. They diverge when lone pairs are present.

My experience with exceptions that break VSEPR

About five years ago, a graduate student brought me a crystal structure of a xenon compound—XeF4—and asked why VSEPR predicted square planar while their quantum chemistry calculation showed slight distortions. The theory says six electron domains with two lone pairs occupying axial positions gives a square planar molecular geometry. The calculation showed bond angles deviating by about 1.2 degrees from perfect 90-degree separation. The lone pairs on xenon are stereochemically active but diffuse, and their repulsion isn't perfectly uniform because xenon's valence shell includes d-orbital contributions that VSEPR completely ignores. The workaround I recommended was running a DFT calculation with a relativistic effective core potential for xenon and comparing the optimized geometry to the VSEPR prediction. For lighter elements like carbon, nitrogen, and oxygen, VSEPR predictions are usually within 2-3 degrees of experimental geometries. For heavier elements in periods 4 through 6, expect deviations of 5 to 10 degrees, sometimes more. The theory is qualitatively correct but quantitatively inadequate for anything past the third period. I've also seen students consistently misidentify the geometry of SF4. Four fluorine atoms bonded to sulfur plus one lone pair gives five electron domains. The electron geometry is trigonal bipyramidal. The lone pair occupies an equatorial position to minimize repulsion. The molecular geometry is seesaw or disphenoidal. Most students guess see-saw from the name alone without working through the domain count. When I make them draw the structure and count domains before naming anything, the error rate drops from about 40 percent to under 10 percent.

When VSEPR fails entirely

Coordination compounds are the biggest failure case. Take [PtCl4]2-. VSEPR would predict square planar for four bonding domains around platinum if you counted d-electrons incorrectly. The actual geometry is square planar, but not for the reason VSEPR suggests. It's due to crystal field stabilization energy and the d8 electron configuration of Pt(II). VSEPR has no concept of d-orbital splitting or ligand field effects. It just counts domains and applies repulsion rules. For transition metal complexes with more than eight valence electrons around the central atom, or for lanthanides and actinides, the theory gives meaningless results. Benzene is another trivial case. VSEPR around each carbon predicts trigonal planar geometry, which happens to be correct. But the theory doesn't explain why all six carbons lie in the same plane or why the C-C bonds are all identical in length. Resonance and delocalization do that. VSEPR treats each bond independently and misses the whole-system behavior entirely. I should also mention that VSEPR cannot predict magnetic properties, spectroscopic behavior, or reactivity patterns. It describes static geometry only. If you need anything beyond atom positions, you need quantum chemistry or at minimum molecular orbital theory.

Practical shortcut that actually works

Here's the method I use now instead of the full VSEPR derivation. Write the Lewis structure. Count the number of atoms bonded to the central atom. Count the number of lone pairs on the central atom. Add them together to get the steric number. Look up the steric number in a table: steric number 2 is linear, 3 is trigonal planar, 4 is tetrahedral, 5 is trigonal bipyramidal, 6 is octahedral. Then remove the lone pair positions and name the remaining shape. This shortcut works because the steric number determines the electron geometry, and the molecular geometry is always a subset of that. You're not skipping any real chemistry. You're compressing the logic into two steps instead of six. Time per molecule drops from roughly three minutes to thirty seconds for simple cases. For exam preparation, this compression saves about twenty to thirty minutes across a typical problem set. The one exception where the shortcut fails is when the central atom has an incomplete octet, like in BF3. Boron has three bonding domains and zero lone pairs. Steric number is three. Electron geometry is trigonal planar. Molecular geometry is trigonal planar. The shortcut still works, but you should recognize that BF3 is electron-deficient and acts as a Lewis acid. VSEPR doesn't tell you that. It only tells you the shape.

Hybridization versus geometry

Students conflate hybridization with molecular geometry constantly. They're related but distinct concepts. Hybridization is a mathematical construct used to explain bonding in valence bond theory. Geometry is an observable physical property measured by X-ray crystallography or microwave spectroscopy. sp hybridization corresponds to linear geometry. sp2 corresponds to trigonal planar. sp3 corresponds to tetrahedral. But sp3d and sp3d2 hybridization involve d-orbitals that contribute negligibly to bonding in most main-group compounds. The hybridization model is useful as a bookkeeping device but shouldn't be mistaken for physical reality. Recent computational studies show that for second-row elements like carbon and nitrogen, d-orbital participation is vanishingly small—on the order of 0.1 percent of the bonding character. Calling it sp3d hybridization for PF5 is technically misleading even though it predicts the correct trigonal bipyramidal geometry. The actual bonding is better described by three-center four-electron bonds or by molecular orbital theory without invoking d-orbitals at all. I mention this because advanced courses and research papers don't use the hybridization model anymore. If you're preparing for graduate work or research, learn the limitations alongside the predictions. Otherwise you'll carry misconceptions into territory where they cause real problems.

Resources for deeper study

The original VSEPR paper by Sidgwick and Powell dates to 1940. Gillespie and Nyholm refined it in the 1950s and 1960s. Their work is readable but dated. For a modern treatment, Housecroft and Sharpe's Inorganic Chemistry covers the theory alongside its failures. Cotton and Wilkinson is the classic reference but assumes considerable background. Online, the LibreTexts Chemistry module on molecular geometry includes interactive 3D models that rotate on screen, which helps with spatial reasoning more than any diagram in a textbook. For hands-on practice, I recommend building physical models with a kit. The tactile feedback of seeing lone pair positions force bonding pairs into specific angles beats any number of practice problems. Students who work with model kits typically score 15 to 20 percent higher on geometry identification questions within two weeks of regular use.

Bottom line on what this theory can and cannot do

Molecular geometry prediction via VSEPR is fast, intuitive, and surprisingly accurate for main-group compounds with steric numbers up to six. It fails for transition metal complexes, heavy p-block elements, systems with significant delocalization, and anything requiring quantitative precision beyond approximate bond angles. Use it as a first approximation, not a final answer. When the theory gives ambiguous or incorrect results, switch to computational methods or consult crystallographic databases like the Cambridge Structural Database for experimentally determined geometries. The geometry of a molecule determines its polarity, its intermolecular interactions, its reactivity, and ultimately its function in biological and materials systems. Understanding shape is foundational. But foundational doesn't mean complete. Treat VSEPR as an entry point, not an endpoint.