[tmartin]

Something else I was thinking of... we're focusing on the H-bonding ability, but with the pronounced solvent effects, could it be possible that even if we put a better hydrogen bonding partner there, that the ratio would only increase to 2:1 favored?  If we made the acid, would it even be soluble enough in hexanes (the solvent giving the best ratio previously) to run the reaction?

I think the addition of a Lewis acid could help, but wouldn't that be acting similar to what the H-bonding would accomplish?  With less of a pronounced solvent effect?

Just to throw out another different idea while we wait for a hint or confirmation from azmanam... could we differentiate another spot on the molecule.  Put another substituent somewhere that would tweak the transition state and make the one we want more favorable, and would be easily removed or incorporated into the natural product?

Maybe formation of the lactone would mess up the arrangement in the TS or cause the loss of some favorable overlap?  

[Heory]

Maybe formation of the lactone would mess up the arrangement in the TS or cause the loss of some favorable overlap? 

I cann't agree more with Mr. Cock, for it's also my idea. I can't express clearly due to my poor English. I will think of another way, according to your hints.

[azmanam]

Reducing the ester would be a good solution and still allow for H-bonding, but I wonder will it significantly reduce the A1,3 strain?

Quite true.  We'd still have A1,3 strain in that case.

I would think if there was a group better able to H-bond than the ester

On the right track, now.

If we made the acid, would it even be soluble enough in hexanes (the solvent giving the best ratio previously) to run the reaction?

Good thinking.  I suspect with the methylene chain, t-butyl group, and furan, that it could possible go into hexanes, but that would certainly be something to think about.  In fact, they did not convert to the free acid.

How about increasing the potential of the ester to H-bond - convert it to a carboxylic acid, amide or carboxylate?

Dan for the win.  Everyone circled the answer, but Dan was the first to mention amide formation.  Amide oxygen atoms are much more Lewis basic than ester oxygen atoms.  The authors did in fact swap the ester for the amide, and ran the reaction again with the free alcohol.  In hexanes, with the new, more Lewis basic amide, the selectivity finally favored the desired diastereomer.  Only a modest 3:1 dr, but overcoming a 1:25 dr in favor of the wrong diastereomer, I think the authors felt OK about 3:1.  Good teamwork and problem solving, all

I’m thinking why the authors didn't make a lactone. They didn't know the method?

Oh, I'm quite sure they knew about lactonization.  In fact, that is how they determined the diastereoselectivity of the 2+2.  After the photocycloaddition, one of the diastereomers is quite susceptible to lactonization, the other is geometrically impossible.  So conversion of one diastereomer to the lactone confirmed the orientation of the stereocenters in that product.


(edited to correct image)

As to why they didn't lactonize before?  I really don't know.  The step after 2+2 is in fact lactonization, and the lactone remains until like 5 steps from the end.  Perhaps the fused 5,5 system can't form with the unsaturation at the fused position?  I don't know if I buy that argument, because I can build a model of the fused lactone, and it doesn't seem too strained.  I can also fold that into the desired transition state with ease and no major steric restrictions.  Tis a mystery, I suppose.

This work comes from Crimmins synthesis of ginkgolide B.  Quite a fun read all the way through, but I thought this conformational analysis problem that also got us thinking about what to do when reactions don't go the way you plan was worth talking about in detail.

[azmanam]

forgot to include the references...

http://dx.doi.org/10.1021/ja001747s

[Heory]