Have you seen the chemical engineering category on the wiki?

Also, I have been interested in 3d printers too for some time, but unfortunately have not had any space or money for experimenting.  Have you seen the reprap Metallica project?  Very interesting.

I have thought and read a lot about this sort of thing and I mean to post my notes and conclusions on the reprap forum when I get around to it.

I think there is a lot of potential for a smaller Metallica.  Just big enough to fit in a large mason jar for example.  The smaller size gives it higher accuracy, and it could print things like piston heads, fittings, connectors, hydraulic pistons, etc. and maybe bearing races or something too (probably need final cleaning before use) out of nearly any desired metal alloy to a few microns accuracy.  And most of it's own parts.

Separately, check out direct laser sintering.  If you used a laser instead as the heat source to avoid the need to have the heated spot grounded, and could both deposit and remove powder effectively.... You could print in a wide range of materials on a single workpiece.  That would really be something. 

Just deposit the powder as normal, sinter in the right spots.  Okay, one material printed.  Now remove all (really all, that might be an issue) excess powder. Repeat, with a different material. 

It is already possible to print in 2 materials to some degree, as printers that use a support material show (I think, need to double check that).

I'm not saying it would be easy to work out the details, but DLS already works, multimaterial done like this would be I think largely a matter of depositing and removing the powders effectively.  Come on, there's gotta be some way to do it.  Feedback for layer thickness might help.  Using the laser a bit like a scanning electron microscope (but with light) might make for a useable way to get high res imaging feedback without expensive cameras - just one , or for stereo, 2 photosensors, which could be printable with the printer.

Another awesome thing is that if you had feedback about the surface under the laser, you could put 2 such machines side by side to some degree, and have them work on the same workpiece.   Scalable, capable of printing large objects like cars straight from commodity powders.

Another issue is the cost per unit power output of the laser, but I think that could be addressed with a printable gas laser, such as a TEA laser of some sort, or just a CO2 or similar low pressure laser, although the wavelength of CO2  lasers is a bit long.  You might not be able to print all of it, but at least most of it.  And the parts you can't print?

Once your printer is capable enough, for anything you can't print, you could probably print a machine to make that thing. Just include the design for any such machines that the printer needs to print itsself from commodity powders, and you now have a truly self printing printer.  There are certain manufacturing methods that increase, not decrease the accuracy of the machining. 

Float glass is one example.  A molten layer of tin is easy to make with low accuracy machining, and extremely flat and smooth.   So I think it will be possible.  Grinding is another example, the motion of the grinding wheel allows it to produce a surface far smoother than the wheel (obviously).  Smoothness and flatness are merely different scales over which shaping accuracy is measured.

Similarly, with high res feedback you can grind, take a look, grind some more, take a look, and so forth so you can move the precision out of the klunky machine and into the sensing part of it, which can be easier.

This of course is the sort of thing that got us from stone axes to fab facilities, so it certainly works.

Just out of curiosity, are you like really really serious about this, or just a hobby type stuff?

here is a link:http://reprap.org/wiki/MetalicaRap

took some time to look over te metalica(one "L" made me miss it).  It's a quite promising system.  It's much simpler than the SLS, but I'm also curious about how fast it could actually go.  Something I've been interested in is bulding a very large(volume measured in feet, not inches) SLS bay.  You could print out say, an engine block.  The EBM system promises to do it without needing finishing work, but even SLS would probably take days to make one, can the beam spread to print larger areas at a time, speeding things up?  I was also looking at things you can't get from a cast/sintered part--in the firearms world, a cast receiver is considered inferior to some extent to a forged part.  The physical act of forging changes the metallurgy.  I was looking initially at basically just putting a heavy roller in to work each layer as you lay it down, providing some of both worlds, but I don't know enough about metal to know if that would be reasonable.

If you can get powdered metal without trouble though, a printer could make blanks for rolling into angle or tube stock, meaning you might get around a lot of the early metal working tools.

I was thinking(and mentioned it on a thread dedicated to the topic) of using an ion beam to "print" ions to build up microchips.  An ideal system would be able to do larger particles, so that a simpler circuit could be built up faster.  One of my brothers and I talked of a controlled ion deposition system--an electron beam would be used to ionize a particular spot to attract a free floating ion in the vacuum chamber, which would neutralize that charge and on to the next.

I did mention COTS semiconductor materials.  Someone on another forum pointed out there are differences in performance between the multicrystal devices made by NanoSolar's process, and the solid ones made by conventional lithography.  If the electron beam can be focused sufficiently to produce the required accuracy on ion placement, the device could make semiconductors easily, up to the modern 40 nanometer standard, depending.

Such a system could be sold to chip makers, who might prefer it, reconfiguring their lines like the companies using FDM machines to do production work.

I like that the electron beam can be made on the device.  The only thing lacking is feedstock.  Nano-scale powder is going to be tricky to make.  The wire system as mentioned with NASA might work better since drawing wire will be easier than making metal powder, and it will be less dangerous to the user--nanopowders are a health risk when inhaled.

If the system could say, pull the aluminum out of moon dust as it goes--print a layer, peel off slag, repeat--it would be a nice thing to have when making a moon base.  Being able to get the feedstocks is the big trick to being able to use things like this.  Plasma recycling can burn almost anything, and get even small amounts of material out of rock, but is beyond the scale of the GVCS.  Perhaps if a Focus Fusion style reactor were added, but I'm waxing sci-fi.  A "solar village construction set" would be interesting, but far in the future.  Even the EBM machine is going to take a fair amount of power though.  Since I'm willing to get on the grid for my first fabber, I'm less concerned with power than OSE is, wanting to be able to make things literally from scratch.