Electron beam lithography is a method for creating extremely small and precise patterns on a surface, either for etching or sketching. The principle is similar to Photolithography, but has several important differences. The major difference is that the exposure is not done all at once with a mask, but rather point by point with a rastering narrow beam of electrons, normally in the 20-100 keV range. This makes it significantly slower - writing a large complex pattern can take hours. Photolithography, on the other hand, rarely takes over a minute. Electron-beam lithography is thus ideal for custom-building situations, such as in the scientific arena. A second difference is that until recently, the smallest possible features obtainable by e-beam lithography were much much smaller than those obtainable by photolithography. E-beam lithography can make features on order of 20 nanometers. This feature size is quite close to some fundamental limits of the e-beam method, so high-end photolithography has been catching up, and in the long run will surpass it.
First, you apply a special kind of plastic called a resist onto the surface, by dropping the dissolved polymer onto the surface and spinning it rapidly to form a smooth layer - 3000 rpm for 40 seconds would be typical. I generally used PMMA.
Then you shine an intense, focused electron beam onto selected regions of the resist with a Scanning Electron Microscope. Depending on the kind of resist used, this will break up the resist (positive resist) or cause it to form stronger chemical bonds (negative resist). For PMMA, a typical dose would be 550 microcoulombs per square centimeter; some resists are much more sensitive.
Then you develop the resist to make the weak parts dissolve. For PMMA, dip in a 1:3 mix of MIBK:IPA for around 8 seconds, and immediately rinse in pure IPA.
Now you have a pattern of resist on the surface, with some areas exposed and others covered by the remaining resist. At this point, there are multiple things you can do.
One option is to etch away the exposed areas. For example, a SF6 plasma rapidly destroys SiNx. So, if your surface was made of SiNx, you can bombard it with an SF6 plasma to etch it away only in the exposed areas. The covered areas will have their resist depleted somewhat, but you can lay enough down that the surface remains covered. The process will depend on what you're etching. There are other plasma etches, and many liquid-dip etches.
Another option is to deposit material (often an evaporated metal) onto the surface. Once the metal is down, you lift off the resist, taking the metal which fell on it off too. All that is left on the surface is tiny wires in the pattern of the exposed areas. In this case, it is useful to have the metal which is to be removed not be in contact with the metal that is to stay behind. To achieve this, one tries to make the resist develop an overhang, known more commonly as undercut.
Undercut can be achieved in two ways. The main way is just to let it happen on its own. Unless your surface is very thin, some secondary electrons will be generated in the surface, and some of them will be reflected back up into the resist. Consider what happens then:
After metal deposition, it has this cross-section:
... m ...
and there is no metal contact. After liftoff, it has this cross-section:
If the electron reflection doesn't provide enough undercut, one can layer an insensitive resist on top of a sensitive one. The lower layer will be more cut away even though it receives slightly less electron radiation.
I already mentioned one common positive resist, PMMA. It comes in several concentrations and with different solvents, useful for getting different thicknesses. Other resists are NEB, EBR, ZEP, and UV. Of these, only NEB is a negative resist. All of these are toxic enough you really don't want to be in contact. Also, while the resists themselves are not at all small molecules, their solvents often are, and those solvents are also toxic. Fortunately, though, neither the resists nor their solvents are corrosive. So, nitrile gloves are adequate protection, while latex are not!