I am a new user of Avogadro and have been using it for energy minimization of molecules.
So, I uploaded a molecule that was drawn using ChemDraw and its initial energy calculated by Avogadro was 8.23 kJ/mol. I then carried out a conformer search for its lowest energy conformer (Extension - Molecular Mechanics - Conformer search) using Genetic Algorithm search and MMFF94 force field. The energy of the molecule after the conformer search was 59.60 kJ/mol. Whereas, when I just carried out geometry optimization (Extension-Optimize Geometry) the energy was 1.46 kJ/mol.
It usually helps to know the molecule in question, but I’ll give some general comments. I’m also going to assume that you’re using MMFF94 for both calculations.
The main difference is that you’re comparing two different things. In a conformer search, you’re usually not fully optimizing the molecule to the minimum. You’re trying to quickly find the most likely conformer. So the method tries a particular conformer and performs a few optimization steps – to get a rough geometry, minimize steric clashes, etc. It then tries another conformation, and repeats. The idea is that you pick a conformation that has the lowest “rough” geometry.
Ideally, you’d fully minimize each conformer. But let’s say that takes 100 optimization steps and you want to check 100 conformers. That’s 10,000 energy evaluations. If instead, I do 50 optimization steps, it’s twice as fast.
What you’re seeing is fairly common. The initial geometry is rough because you imported it from ChemDraw (2D). You performed a conformer search (great!) and then optimized the geometry afterwards.
You should have a pretty good geometry at this point.
BTW, take the energies from MMFF94 and other force fields as very approximate. Even approximate quantum methods are significantly better, albeit slower.