The solvents used in flash chromatography are generally polarized, small, and cheap. Rfs are primarily driven by structure-independent electrostatic interactions (H-bonds, dipole-dipole, etc.), and solvent specific steric interactions rarely come into play . This means that solvents (mixtures) of equal polarity are interchangeable, making the solvent system of a column somewhat arbitrary. In industry, solvents are generally selected for a mixture of low cost and low toxicity, while academic labs are more influenced by ease of use and idiosyncratic factors .
In my work, three binary mixes have been sufficient for the large majority of compounds:
1) Hexanes/Ethyl Acetate
The workhorse for most student chemists, this mixture is very good for non-polar compounds that are stable to at least 80℃. Effective mixtures range from 0.5% EtOAc all the way to pure EtOAc, with a 7:3 mixture inexplicably giving an Rf of around 0.3 for at least half of my compounds.
This system is more effective for polar compounds and compounds with poor solubility in ethyl acetate. Most compounds lacking a free amine work well in the 2-20% MeOH range, which corresponds roughly to 20-60% EtOAc in hexanes. Mixtures containing less than 10% MeOH should be prepared fresh each column volume, as methanol and dichloromethane are azeotropes. Evaporation of dichloromethane will rapidly reduce the effective concentration of MeOH if mixtures are left to stand.
3) Petroleum Ether/Diethyl Ether
Both of these solvents boil at or around 40℃ at atmospheric pressure, which makes this a perfect system for substances too heat sensitive for the rotovap. Both also have high vapour pressures, and can be either removed at atmospheric pressure with a streme of nitrogen, or rotovapped with a room temperature water bath.
A standard, constant binary solvent works very well when two compounds of roughly similar polarity need to be separated. But, it’s inefficient when there’s a large difference in Rfs (ex. 0.3 vs 0.9), an dhas poor resolving power when there are four or more compounds in the sample. Solvent gradients are the answer.
The quick and easy version is to run your first column volume at an extremely low polarity, such as 98:2 hexanes:ethyl acetate or pure dichloromethane. Your compound of interest shouldn’t move, but most common non-polar contaminants will fly through (ex. Excess equivalents of protecting groups). From there, run one CV at a mixture that gives your compound of interest an Rf of 0.3, and one CV at Rf=0.5. By the end of the third CV your column should be finished .
The more methodical gradient column approach is to always use mixture in which your least polar compound has an Rf of 0.3. Designed for identifying low-concentration side products, this will probably require an hour or more and forty or so test tubes, but will allow a detailed understanding of your reaction.
Acids and bases tend to move through flash silica poorly, often eluting in frustrating streaks fifteen or more fractions long . This stems from their ability to transfer protons back and forth across either the flash silica or other compounds on the column. As would be expected, the resulting ions have negligible Rf values in all but the most powerful of solvents, and are essentially fixed in place until a subsequent proton transfer.
The solution is to add in a small (<2%) quantity of sacrificial acid or base, as the case may be. These act by preventing deprotonation/protonation of your compound of interest, and are essential whenever your compound has a free carboxylic acid or primary/secondary/tertiary amine.
Acetic acid is the most common proton source, while either ammonium hydroxide or triethylamine serve as bases. Of the latter two I prefer triethylamine, for its facile solubility in non-polar solvents. All three are easily removed under high vacuum, and have only a minor impact on TLC staining solutions .
 Aromatic solvents like toluene can show minor Rf variations as a result of pi stacking, etc., but I’ve never heard of the difference being significant.
 Ie. “that’s how the last grad student/post doc did it.”
 This 0.3 to 0.5 switch is designed reduce the number of fractions containing your compound of interest. If you are confident that no more-polar impurities are present I’d recommend switching to a 0.5 Rf once your compound starts eluting, whether or not you’re running a gradient column.
 Note: A streaking compound does not mean that it is an acid or a base. Silica is weakly acidic, and suscetible compounds may decompose on the column (or TLC plate). If you suspect compound degradation run a 2D TLC plate (off diagonal spots indicate decomp.), and add ~2% TEA to the column when slurry packing.
 Acids/Bases are just as important in the TLC pot. Triethylamine will stain in ninhydrin, but generally the background is faint enough that there’s no trouble locating your amine of interest. I haven’t extensively tested ammonia, though if you are having trouble it could likely be removed by heating the plate prior to dipping in the staining solution (that works for pyridine, at any rate).