Dominant type seen today. Original ones were true metal gate on an oxide as the name implies, but modern ones tend to use polysilicon instead.

Classical metal gate type. From left to right:gate, source, drain. Note that gate and drain are more or less identical in CMOS. Colors are from thin film interference indicating different layer thicknesses.

Like above, except metal removed. Black gate area is probably salt from dissolved gate metal. Color difference is because its from a different chip, not etch process (although etching can also change color)

Contacts from left to right: collector, emitter, base. The collector forms the bulk of the well. Assuming its an NPN (which it may or may not be), you can also see the p-isolation in blueish-green around the edges. The base forms the larger white area around the emitter. Traces of the metal contacts can be seen as darker regions on the collector's pink area.

Discrete bipolar transistor from TO-3 package.

Another bipolar description: http://zeptobars.ru/en/read/ULN2003-per-element-die-annotation

Diffusion transistor. Does not use photolithography. Above image shows the mesa shape versus other transistor which were more or less flat.

Name comes from the physical mesa shape. Indications that this is because of the component geometries, but also read something that indicated its because they used acid (probably HF) to etch it out and it gets the shape because it etches more on the top by the time it gets to the bottom. Modern dicing is either through scribing or sawing.

Uses photolithography to print transistors. Has SiO2 passivation. Significantly reduced cost and significant increase in quality.

Wafer is put into a high temperature oven and doped by gas diffusing into the silicon.

“”“ It is the world's worst transistor, you couldn't sell it as a discrete component: low cutoff frequency, very limited current range and an inferior noise figure. But no self-respecting analog IC designer would want to be without it. The reason: In either a CMOS or bipolar process no additional diffusions are required. ”“” -“Designing Analog Chips”, Hans Camenzind