Why Your Structures Keep Coming Out Wrong
I spent years correcting student drawings where the electron geometry matched perfectly but the molecular geometry was wrong, or vice versa. The two get confused constantly, and once you mix them up, everything downstream collapses. Let me walk through how this actually works in practice. Electron geometry describes the arrangement of all electron domains around a central atom — bonding pairs, lone pairs, and single electrons count equally. Molecular geometry describes only the positions of the atoms themselves, ignoring lone pairs. They diverge the moment you have one or more lone pairs on the central atom. That's it. That's the core distinction. Everything else is just applying it correctly. Here's the method I use. Count total valence electrons. Draw the skeleton. Place bonds. Distribute remaining electrons as lone pairs. Count electron domains around the central atom — that's your electron geometry. Then look only at where the atoms sit and name that molecular geometry. Repeat. It takes about 90 seconds for simple molecules once you've done it a dozen times.
The most common mistake I see is people counting regions of electron density and then forgetting that lone pairs occupy space but don't show up in the molecular shape name. Tetrahedral electron geometry with one lone pair gives trigonal pyramidal molecular geometry, not tetrahedral. That's the error that cascades through every problem set. I ran into this problem with hypochlorite, ClO-. Students would draw it and call it linear because there's only one bond, which is obviously wrong for a diatomic species but you'd be surprised how many do this. The fix is simple: two electron domains, tetrahedral electron geometry (though some texts call it linear for diatomic — that's a notation choice that doesn't matter for VSEPR), and since there's one bonding pair and three lone pairs on chlorine... wait. Diatomic molecules are an edge case here. For one atom bonded to another, the molecular geometry is always linear by definition because two points define a line. The VSEPR model doesn't add much value for diatomics. Don't waste time on them. More useful are things like XeF4. Square planar molecular geometry, octahedral electron geometry. The two lone pairs sit opposite each other because that minimizes repulsion. That's the VSEPR rule you need to memorize — lone pairs prefer 180-degree separation. Ignore that and you'll draw cis configurations when trans is correct.
Another pitfall: resonance. When you have resonance structures, the electron geometry stays the same across all forms. Some students redraw the geometry for each resonance contributor and get confused. It doesn't change. Delocalization affects bond order, not the domain count. The biggest limitation of VSEPR-based geometry prediction is transition metals. For d-block complexes, crystal field theory and ligand field theory matter far more than simple electron domain counting. Octahedral, tetrahedral, and square planar geometries for metal complexes don't follow the same simple rules because d-orbital splitting energy competes with pairing energy. If you're dealing with coordination compounds past the first row or with heavier metals, VSEPR is unreliable. Use crystal field diagrams instead. It takes longer initially but gives you actual predictions rather than guesses. For main group elements, VSEPR works about 85% of the time for textbook problems. The 15% failure cases usually involve molecules with significant ionic character, very large central atoms where steric effects dominate over electronic effects, or radicals where the single electron creates unusual geometry. Iodine heptafluoride, IF7, is a classic headache. Pentagonal bipyramidal electron geometry, but the bond angles don't match what simple models predict because fluorine's electronegativity pulls electron density away and distorts the ideal angles. Again, VSEPR gives you the framework but the actual angles deviate.
Get the Full Details
If you need computational geometry optimization for anything beyond simple molecules, check out Open Babel. It's free, handles thousands of molecule types, and outputs optimized geometries in multiple formats. The command line version processes a standard organic molecule in roughly 30 seconds on modern hardware. For quick academic work, that's usually sufficient.
Practical Tips That Actually Matter
Draw the Lewis structure first and never skip it. A wrong Lewis structure guarantees a wrong geometry. I've seen people skip this step and still get the right answer by accident, but you can't rely on that. It takes maybe 30 extra seconds and saves you from completely wrong conclusions. When counting electron domains, treat double and triple bonds as a single domain. This trips up everyone at least once. A C=O bond is one region of electron density, not two. The pi bond doesn't create an additional domain for geometry purposes. Use a molecular model kit or a decent 3D visualization tool if you're struggling with spatial reasoning. Physical manipulation of the geometry makes the relationship between electron domains and atomic positions click faster than any amount of reading. I recommend this for students who find the mental rotation part difficult.
For polyatomic ions, remember to adjust the electron count for the charge before drawing the structure. Miss that step and your entire domain count shifts. I've lost count of how many times students forget the negative charge on carbonate and end up with the wrong number of electrons, then the wrong geometry, then the wrong dipole moment conclusion.
