Getting Real About Molecular Geometry
Most textbooks introduce this stuff by saying electron pairs repel each other. That's technically correct but about as useful as saying water is wet. The actual mechanic is more nuanced, and if you're trying to predict geometries for anything beyond simple molecules, the simplified version will lead you astray pretty quickly. Here's how I actually approach it in practice. You start by drawing the Lewis structure, then you count electron domains around the central atom. A single bond counts as one domain. A double or triple bond also counts as one domain. A lone pair counts as one domain. The total number of domains gives you the basic geometry: two is linear, three is trigonal planar, four is tetrahedral, five is trigonal bipyramidal, six is octahedral. Then you look at how many of those domains are bonding pairs versus lone pairs to figure out the actual molecular shape. That's the standard algorithm. The problem is that students and even some working chemists treat it like a rigid set of rules rather than a qualitative model with real limitations built in.
Practical Application of Valence Shell Electron Pair Repulsion Theory
Let me walk through something that trips people up regularly. Take XeF4. You draw the Lewis structure, count your domains, and on paper it looks straightforward. Xenon has eight valence electrons. Four go into bonds with fluorine. That leaves four electrons, which is two lone pairs. Six electron domains total, which means octahedral electron geometry. Two lone pairs go opposite each other, so the molecular geometry is square planar. Standard textbook answer. But here's where the theory gets uncomfortable. The VSEPR model assumes that all bonding domains are equivalent and that lone pairs just push everything away uniformly. In reality, the lone pairs on xenon aren't sitting there symmetrically in a way that the simple model captures. The actual electron density distribution is more complex, and computational chemistry shows that the lone pairs have significant s-character while the bonding pairs are pushed toward p-character. The model gets the right answer but for the wrong reasons if you think about it too hard. I ran into a specific problem last year when someone asked me about BrF5. The textbook answer is square pyramidal, four bonding domains plus one lone pair on bromine. Easy. But when we actually looked at the X-ray crystallography data, the axial fluorine came in noticeably shorter than the four equatorial fluorines. The VSEPR model predicts they should all be roughly equivalent in their behavior relative to the lone pair. They're not. The bonding pairs are being compressed by the lone pair to different degrees depending on their orientation, and the model doesn't really quantify that.
The workaround I use is to remember that lone pairs occupy more angular space than bonding pairs, but the amount of space varies by the type of atom and its oxidation state. For heavy central atoms with high oxidation states, the lone pairs tend to be more diffuse and actually exert less repulsive force than the simple model suggests. This is why you'll sometimes see geometries that look almost exactly like the prediction without any lone pair, even when the math says there should be one affecting things noticeably. Another thing that catches people is the case of molecules with three electron domains where one is a lone pair. You'd expect the bond angles to be somewhere below 120 degrees, like ammonia below 109.5. That's usually right, but not always. In SO2, the bond angle is about 119 degrees, barely compressed at all from the ideal trigonal planar geometry. The reason is that the double bond counts as a single domain but carries more electron density, so it pushes harder against the other domains than a single bond would. VSEPR does account for this qualitatively, but most people gloss over the part where multiple bonds get extra repulsive weight. Here's the part that almost no introductory course emphasizes: VSEPR breaks down for transition metal complexes. You can't just count d-electrons the same way you count s and p electrons. The d-orbitals have very different spatial distributions, and crystal field theory or ligand field theory is what you actually need. People try to apply VSEPR to things like [PtCl4]2- and get the right answer by accident, but the reasoning is wrong and it falls apart as soon as you deal with anything involving pi-bonding or significant d-orbital participation in the geometry.
Get the Full Details
For main group elements, the model works well enough for common oxidation states. Beyond that, you're relying on an approximation that was never meant to be quantitative. If you need actual bond angles and bond lengths, you run a DFT calculation or pull data from the Cambridge Structural Database. VSEPR is a teaching tool and a quick heuristic, not a predictive engine with error bars. The biggest practical mistake I see is applying the model to molecules with delocalized electrons without thinking about it. Benzene has six electron domains around each carbon, all bonding, so VSEPR says trigonal planar. That happens to be right, but the model isn't actually describing the pi system at all. It's describing sigma framework geometry, and nobody usually clarifies that distinction. When you have resonance structures, you pick one and apply VSEPR to it, but the real molecule is a hybrid and the electron density isn't what any single Lewis structure suggests. I also want to flag the case of hypervalent molecules like PF5. Five domains, trigonal bipyramidal geometry. The model works fine here, but the explanation involving d-orbital participation in bonding is controversial at best. Modern computational chemistry suggests that the bonding is better described using three-center four-electron bonds, and the d-orbitals contribute very little. The geometry prediction is correct, but if your instructor is asking about the bonding mechanism, the simple VSEPR explanation is incomplete.
One more edge case worth mentioning. When you have a central atom with both a lone pair and highly electronegative substituents, the bond angles can shift in ways that seem contradictory. Take OF2 versus H2O. Both have the same electron domain geometry, but the bond angle in OF2 is about 103 degrees while water is about 104.5. The more electronegative fluorines pull electron density away from the oxygen, which actually reduces the repulsion between the bonding pairs and allows the lone pairs to compress the angle more. This is sometimes called the Bent's rule effect, and it's something the basic VSEPR model doesn't explicitly include but experienced people factor in mentally. If you're studying for an exam, the algorithm I described gets you through about 90 percent of the problems you'll encounter. The remaining 10 percent is where you need to remember that the model is qualitative, that electronegativity matters, and that lone pairs on heavy atoms behave differently than the textbook diagrams suggest. Beyond that, you need actual computational or experimental data because the theory stops being useful.
