Detecting Dislocations in Trenches by Chemical Etching

With the development of the 4 Megabit Dynamic Random Access Memory (DRAM) and of the eighties, a new process was introduced into Si technology: Trench etching for trench capacitors.

A trench, contrary to the literal meaning of the word, is simply a small hole - typically 1 µm in diameter and (6 - 8) µm deep. It is made to provide a large area for the capacitor while still using only about 1 µm² in "real estate" on the chip surface.

All kinds of problems were encountered with the performance of the trench capacitors, some - maybe - caused by dislocations ending inside the trench.

Since there are about 4 · 10⁶ trenches on one chip, and some 100 chips on a wafer, looking with an electron microscope at a few trenches will not do much good - you will turn to defect etching,

There is a clear question to the analytical people then: Are there dislocations inside the trench, and if yes, does his correlate with electrical performance? Well, Wendt, Sauter and Kolbesen of Siemens AG answered this question in an elegant, if tricky way with chemical etching. Here is what they did.

If you etch your wafer with some defect etch, you may obtain pictures like the following one

This is a SEM picture, because you would not see very much with a light microscope - the big holes being the trenches are just about 1 µm across. In fact, while you would detect the relatively shallow dislocation etch pits marked by "F", you would miss the sharp little holes marked with "S".

Well, you see that there are dislocations ending at the surface. What you do not see is if there are dislocations ending on the surface of a trench; i.e. inside the hole. If you look real close, you might on occasion find something as shown below:

OK, here you have the etch pits of a dislocation that starts at the surface at "A" and ends obviously inside the trench at "B" - but still rather close to the surface.

How about deeper down in the trench? How do you look inside a 1 µm trench (with any method)?

Here is the solution:

Etch the whole wafer, producing etch pits inside the trench if there is a dislocation.
Coat everything with a thin layer of S_i3N₄.
Etch off all of the Si, leaving only the S_i3N₄ layer intact.
Inspect the S_i3N₄ layer. It is a kind of "negative" of the trench structure which now is easily inspected.

Pretty tricky (and by far not as easy as it sounds). Here are some results (the "stripes" are artifacts from image processing)

The inverted trench structure - pretty perfect in this case.

The structure marked "V" clearly results from a dislocation running from a trench to its neighbor.

The structure marked "SF" actually shows a stacking fault.

Same thing here - a prominent stacking fault (and a dislocation).

So we have stacking faults, not just dislocations at work here! This is a major finding, that would have been practically unavailable with other methods.

It is a major finding, because now we have a pretty good idea where the defects are coming from: We most likely deal with a new kind of oxidation induced stacking fault, and that give us a clear idea of what needs to be done to get rid of those defects.

With frame

6.1.1 Observation of Dislocations and Other Defects

Oxidation Induced Stacking Faults in Silicon