Organic synthesis as we know it can't go on without metal-catalyzed bond-forming reactions. There are too many of them, and they're just too useful. Palladium's the workhorse, followed by copper, then you've got rhodium, nickel, and a host of others (gold's been popular the last few years). We have a. . .fairly good idea of what's going on in these reactions, but not quite good enough. If we really understood all the factors involved, there wouldn't be six garbonzillion different sets of conditions for these things, would there?
A short paper's just come out in Angewandte Chemie that illustrates some of the trickiness involved. Carsten Bolm's group at Aachen has published several interesting iron-catalyzed coupling reactions using good old ferric chloride. These are aryl-amine, aryl-ether, aryl-amide and aryl-sulfide-forming procedures, which cover a lot of ground. (Interestingly, it was one of those sulfide papers that was recently plagiarized by another set of authors). But there were always a few kinks, such as variable yield depending on which bottle of ferric chloride was used.
Well, organometallic chemists are used to that sort of thing. But Bolm has gone back to look at things closely, in collaboration with Stephen Buchwald of MIT (whose group has published many similar couplings with other metal systems), and found a surprise. The iron chloride isn't doing a thing. In fact, as you go to more and more pure sources of the reagent, the yield drops off. But it never goes away, even with the 99.9% pure stuff. That's because it seems to be copper (I) contaminants doing the coupling, even at the parts-per-million level.
There are some startling tables in the paper. For coupling pyrazole onto an aryl iodide, for example, Bolm's group had found in 2007 that they could get 87% yield using >98% ferric chloride from E. Merck, along with dimethylethylene diamine as a cosolvent. If you use the >98% from Aldrich under the same conditions, though, you get 26% yield. And the Aldrich >99.99 stuff gives you only 9%. But if you add five ppm copper (I) oxide to that last reaction, the yield goes up to 78%. And if you leave the ferric chloride out completely, and just go with a trace of copper, the yield is exactly the same (it goes down if you run the same trace-of-copper without the diamine, which seems to be complexing it up into solution).
The other couplings that were reported seem to follow the same pattern. This must really be a disappointment to Bolm and his group, because their work was, among other things, an attempt to get away from copper and palladium. Still, this appears to be a much cleaner and more efficient copper reaction than a lot of the procedures out there.
This sort of thing has happened before in organometallic chemistry. There's a well-known example of nickel contamination in a chromium-mediated reaction from the mid-1980s, and more recently, a report of supposed "metal-free" couplings which appear to have been catalyzed by parts-per-billion levels of palladium found in the sodium carbonate being used as a base, of all things. Tricky things, these metals.