Of course I got curious.
Here is what the chip looks like when all the insulating stuff gets stripped off its face. Die size is 0.3 x 0.4 mm. I think the chipped areas at top are due to rough handling with forceps. This was the first one I did, and I had not yet figured out how to handle these elaborate little grains of sand.
With a closer crop, we can clearly see some unusual structure. This chip face shows a few of the axis-aligned wiring and circuit elements that I was expecting, but most of the area is covered by a slightly diamond-shaped grid that is offset about 2 degrees clockwise from the X axis, but 2 degrees counter-clockwise from the Y axis. There is also a square axis-aligned grid over on the right side.
I had no idea what that odd grid would be, so I wrote email to RF chip expert Mike Wyatt (mawyatt) to see what he could tell me. A reply came promptly as follows:
Anyway, the small structures are likely what we called “fill”. Fill was used to help with planarizing the surface, especially in areas of high density, small features, but also in large somewhat open areas without much structure. These large open areas are prone to “scalloping” during multiple layer processing, and cause issues with the processing and metallization.
As the CMOS node feature size drops, so does the need for absolute wafer level planarity, and I guess (don’t know) that the offset “fill" angle also helps since most of the underlying features are generally orthogonal with the Si wafer crystal structure which may cause a slight difference in surface hardness during wafer surface grinding, and the offset angle helps relative to the surface grinding angle making a more uniform planar surface.
These fill patterns also help hide the underlying structure from IP theft, and make copying the chip more difficult.
So there we have the explanation -- the chip is doing something under the covers, but we can't see any of it!
When I first showed the above pictures to a friend who has decapsulated some chips, his first comment was something like "Wow -- you not only decapsulated it, you delaminated the center!"
I thought that seemed very reasonable, since in fact I had decapsulated this chip by heating the whole assembly to a dull red and keeping it there for 1/2 hour, thinking to burn off all of whatever might be on it.
But then, wanting some more insight, I disassembled another one using a more gentle approach with chemicals (Goof-Off to start, then nitric acid to move things along), finishing with an ultrasonic cleaner.
Here is the end result of that process, shown as crossed-eye stereo:
You can easily see in the stereo that, after chemical treatment, the chip face is exposed in the central area, but the ends are covered by some sort of opaque glassy material. Before the nitric acid and ultrasound, those areas were covered by what appeared to be gold-plated electrodes, which in turn were bonded to the aluminum antenna.
So it seems that what the first picture actually shows is not that the center has been delaminated, but rather that the surface was covered by two different materials which burned away slightly differently.
Following are some snapshots of disassembly steps.
The original label, suitably anonymized, before any disassembly, front and back sides.
Closeup of the area covering the chip, with only a soft adhesive layer removed using solvent. Note the two small divots over the corners of the chip. Those are where the aluminum eventually contacts the chip.
Underside of the same area, 1 mm tick marks.
The chip showing its contacts, after disruption of the gray plastic with Goof-Off and some scraping with a #2 insect pin.
Photo technique was a mixed bag. The first two high mag images were using Mitutoyo M Plan Apo 20X NA 0.42 on Raynox DCR-150, photographing through 0.17 mm cover slip which dry-mounts the isolated chip to a microscope slide for convenience in handling. The stereo pair is with Mitutoyo 10X, shooting a naked chip which is glued to a microscope slide using washable Elmer's School Glue. (I figured my chances were nearly zero of getting a clean job on the first try. It took three tries, cleaning the chip with ultrasound after I got glue on the face I wanted to photograph. Third try I still got a little glue on the face, but only at the back side where I didn't care enough to do it again.) The whole labels were traditional closeup, and the high mag shots below those were with cell phone through a zoom binocular scope as I worked to get the assembly apart.
If anybody knows how these things are manufactured, I would love to learn how it's done. I assume that the chips start off being manufactured en masse on large wafers. That part I can understand, sort of. But then how they get cut apart and assembled into these labels that sell for pennies apiece, that seems like magic.
--Rik