cross-sections of MLCCs (multi-layer ceramic capacitors)

[rjlittlefield]

Just to set the stage, here's what we're talking about.



These are small ceramic capacitors that sell for about $0.025 each in retail boxes at Amazon. Across a couple of kits, they range in rating from 10 picofarads up to 10 microfarads -- a nice even factor of 1 million.


Inside the yellow outer coating, there's a small rectangular chunk of ceramic and metal, organized as alternating layers, with every other metal layer connected to one lead. There's a nice video on their manufacture at https://www.youtube.com/watch?v=5_OSxBLlcKQ , "How We Make Capacitors | Ceramic" by KEMET Electronics. Further information is provided at https://www.youtube.com/watch?v=gFEYuaY35Vo .

Here you're seeing a cross section of the 0.1 µF size. I count 55 layers of metal, with an average spacing around 10 microns on center.



Below is the corner of a larger one, 10 µF. Spacing is about 3 microns on center. In the full thickness, I count about 320 metal layers.



From what we can see above, I can't figure exactly how the scaling works. Capacity should be proportional to the number of layers, multiplied by each layer's area, divided by the thickness of the ceramic insulating layers, multiplied by whatever the dielectric constant is. There's 100X between 0.1 µF and 10 µF, but in the above images I can only see a factor of maybe 30-40X, from counting layers and guestimating the thickness. Oh well, a bit of mystery is a good thing.

The above images all show the canonical cross section, sliced straight down through all the layers.

That slice is convenient for explanation, but it wasn't quite the way the investigation actually played out.

In fact, the very first capacitor that I got a clean face on was a 4.7 µF that looked as follows.



Here's a closer view. Yeah, diagonal layers...?



The above structure is not at all what I was expecting to see, and I puzzled over it for an embarrassingly long time. Finally I figured out the trick. That particular capacitor had gotten assembled with the ceramic block rotated about 90 degrees, so that instead of slicing through the layers I had sliced not quite parallel along them. The strange appearance is due to the angle of the slice, not the structure of the chip.

Nearing the end of the exercise, of course I had to try sectioning the smallest cap I had, only 10 picofarads. In that case I was expecting to see just a couple of plates, with quite a large distance between them. But to my surprise, this is what appeared:



Gentle probing with an ohmeter and sharp pins shows that both of the angled plates on the left side are connected to the left lead, while both angled plates on the right side are connected to the right lead. Why there are two of each, and why/how they are manufactured at that angle, at this time I have no idea.

Notes on technique... The capacitors were glued to glass slides with Krazy Fix Light Cure Super Glue, then progressively ground down with a 1200 grit diamond stone followed by 7000 and 15000 grit sandpaper. I assume that anybody with decent lapidary skills could put a fine polish on these things in no time, but I was working by hand and stopped when they were good enough to see the structure I wanted. The appearance was slightly improved by wiping on a thin layer of "Enco No.1 cutting fluid" just before photographing. I tried glycerin first, but that wouldn't stick at all -- just beaded right up. I have no idea what's in the cutting fluid, except the decades-old label says "a synthetic cutting compound formulated for all ferrous and non-ferrous metals". Photographed with 5X, 10X, and 20X Mitutoyo objectives in the usual setup with Raynox DCR-150, Canon R7, and LED tube diffuser with Jansjö LED ilumination.

I am amused to note that I cannot recall why I decided to look inside these things. As always it took much longer to do the work than to have had the idea.

That choice may have been prompted by some of the illustrations in a very nice book that I recently purchased: "Open Circuits: The Inner Beauty of Electronic Components", https://www.amazon.com/dp/1718502346 . If you're curious about electronics, I heartily recommend it.

--Rik

[Beatsy]

That's a fascinating little project with very interesting results Rik. Neat!

[WojTek]

Hello Rik,
Great idea, interesting photo subject, fantastically photographed and prepared with love :-)
Best, ADi

[rjlittlefield]

I'm glad you found it interesting.

I certainly was not expecting that last step of strange structure with the 10 pF size. My brain is now toying with the idea of some sort of automated sand-and-shoot setup that could make a time-lapse of sectioning its way through the whole structure. Probably that'll never happen, but who knows?

Along more conventional lines, here's a stereo view of what I got when I used blunt force to get a quick look inside a 4.7 µF unit:



I don't think you can tell by looking at the image, but this stereo shows exaggerated depth produced by 18 degrees total separation (plus and minus 9 degrees). I used my newly acquired 20X NA 0.75 objective with a cover slip glued to it, which gave me a nice thin focus band (nominal 0.8 micron) to resolve depth in what is physically a pretty flat subject.

The whole stack was 330 frames shot at 0.5 micron, so total depth about 0.165 mm versus frame height about 1.1 mm. Processed here as every 3rd frame, at 50% scale, using DMap with a 3-pixel estimation radius. Usually I use PMax for stereo, but with this subject that produced very distracting streaks around the bright metal reflections.

--Rik

[WojTek]

Hello Rik,
Yes, the depth on the stereo photo is quite large.
Best, ADi

[rjlittlefield]

the depth on the stereo photo is quite large

Well, it's easily seen. But in terms of Zerene Stacker's controlling value, stereo shift for the above image is only plus and minus 3.6%, barely larger than the default value of plus and minus 3.0%. If I had not told you that the shift percentage corresponds to plus and minus 9 degrees, I doubt that any oddness would have occurred to you.

Here is another view of the same subject, shot at lower magnification and in a much different physical position. This is only 5X NA 0.14, with the subject rotated roughly 30 degrees backward so that we can see the overall shape. In this case the stereo shift is only plus and minus 3 degrees, for a total of 6 degrees between eyes.



By the way, the different renderings of gray scale and contrast are mostly not due to post-processing. The 20X NA 0.75 stack permits only much lower angles of illumination, and in addition most of the subject is much closer to "flat" as in perpendicular to the optical axis. I think that the combination of these factors, combined with the fine granular surface texture of the ceramic, causes much more of the surface to be in deep shadow with the higher mag objective.

--Rik

[WojTek]

Hello Rik,
OK, your comment may have influenced me.
However, the current photo also offers an excellent stereo effect.

I think that the combination of these factors, combined with the fine granular surface texture of the ceramic, causes much more of the surface to be in deep shadow with the higher mag objective.

Maybe the surface texture could be better visualised on the 3D model.
Best, ADi

[ray_parkhurst]

I did a lot of this kind of work in my last job at a MEMS startup (Cavendish Kinetics, now owned by Qorvo). Quality was one of the many hats I wore at that company. Most of the work was looking at failure analysis of the packaging, but occasionally I did some actual MEMS xsections, and I did some slice-by-slice animations as well to show problems and such. Unfortunately I can't share any of them.

[CrispyBee]

I'm glad you found it interesting.

I certainly was not expecting that last step of strange structure with the 10 pF size. My brain is now toying with the idea of some sort of automated sand-and-shoot setup that could make a time-lapse of sectioning its way through the whole structure. Probably that'll never happen, but who knows?

Along more conventional lines, here's a stereo view of what I got when I used blunt force to get a quick look inside a 4.7 µF unit:



I don't think you can tell by looking at the image, but this stereo shows exaggerated depth produced by 18 degrees total separation (plus and minus 9 degrees). I used my newly acquired 20X NA 0.75 objective with a cover slip glued to it, which gave me a nice thin focus band (nominal 0.8 micron) to resolve depth in what is physically a pretty flat subject.

The whole stack was 330 frames shot at 0.5 micron, so total depth about 0.165 mm versus frame height about 1.1 mm. Processed here as every 3rd frame, at 50% scale, using DMap with a 3-pixel estimation radius. Usually I use PMax for stereo, but with this subject that produced very distracting streaks around the bright metal reflections.

--Rik

That's a really impressive photo!
Never thought about the intricacies of capacitors, interesting use of the cutting fluid! You basically used it like a glaze or varnish, I like that idea! I've often wondered whether it would be a good idea to coat specimens/objects with fluids like oils to enhance the surface contrast (especially for cross-polarisation), The results look really good!!

[rjlittlefield]

Thanks for all your comments and questions. I apologize for my delay in responding! In doing some further investigation I ended up diving into a rabbit hole, which has been consuming most of my think time for the last week or so.

I've often wondered whether it would be a good idea to coat specimens/objects with fluids like oils to enhance the surface contrast (especially for cross-polarisation)

Wetting the surface is a standard trick that works really well sometimes. At the least it often gets rid of a lot of diffuse surface glare that kills contrast and hides colors. As a child I was taught to lick rocks ... uh, get them wet ... to see the true colors. It can also be used to smooth out imperfections on transparent surfaces that you want to shoot through. My own classic example, and a possibly amusing story, is the Christmas LED at https://www.photomacrography1.net/forum ... php?t=4089 .

But the most impressive example that I've seen recently is in a National Geographic episode of "Photographer", featuring Anand Varma struggling to get a decent video of a chick hatching. If you have a Disney Plus subscription, then at https://www.disneyplus.com/play/a1ec72d ... 180726583f , at 41:30, you can see (finally, in desperation) oil being dropped on the membrane inside a partially opened egg (the last one, according to the storyline). The membrane suddenly becomes clear, and the closed captioning says "Woah, what? That's incredible! ... Turns out, drop mineral oil on the membrane and you can see everything that's happening." The visual effect really is striking.

In this thread, I only wetted the sanded surfaces. For the fractured surface, I didn't want anything messing up the fine structure so I left that dry.

That's a really impressive photo!

The 20X stereo did come out nice, but only at the resolution posted. When I zoomed in tighter, it was a lot less sharp than I expected.

The reason turned out to be that when I shot the stack, I had accidentally stopped down the objective by quite a lot when I bumped the handle of the adjustable iris that I use to increase DOF to facilitate specimen positioning. The result was that my lovely NA 0.75 objective was being operated a lot narrower than that.

So, I reshot that stack with the objective full open. The result then looks good even at actual pixels. You can access a camera-resolution stereo pair HERE.

Here is a crossed-eye stereo pair of an area that I thought was particularly interesting. (Click image for universal LRL format.)



This is plus and minus 5 degrees, which does not exaggerate the perspective. That separation is exactly the same as my stereo microscope and also corresponds to viewing at 13.5 inches distance with standard 60 mm between eyes.

I'm not sure what the lumps of white debris are, but my main interest is the exquisitely thin ridges and spikes that are formed by the metal film.

Those ridges and spikes are easily seen, all over the frame, when viewing in stereo. But they are very difficult to see or figure out in any single static image.

Rocking works OK in some cases:


I can see the 3D structure of the spikes in this rocking view almost as well as in the stereo view.

But for me the rocking view is uncomfortable because it's always moving and there's no way to make it stop.

Rocking also has the huge disadvantage that it only works well when the center of rocking is located very near the thing whose structure I want to see.

In this case, I deliberately made that happen for the spike just left of image center, by adjusting the left/right positions of the two images so as to leave the base of the spike in one place while everything in front or behind that moves.

But that same positioning works less well for other parts of the scene. In some areas that are only slightly outside this frame, rocking with the same positioning totally fails because all I see is something of indeterminate shape moving left and right.

This is very frustrating. One of my big goals in photography is to show somebody else the same structure that I can see, and for a subject like this and a viewer who cannot see stereo, I'm struggling to get even close to that goal.

Maybe the surface texture could be better visualised on the 3D model.

That may be, and exploring that possibility is most of what I've described as a rabbit hole.

Skipping past all the difficulties for now, my best current result can be seen at https://sketchfab.com/3d-models/microst ... b2e68507a9 , "Microstructure of a fractured MLCC die".

Here is what that looks like in standard render and with Matcap rendering:




Then, by rotating the view and using the mouse to rock the view as diagrammed here, I am able to figure out the structure pretty well. Once I have it set up, the rocking is comfortable because it's under immediate control of my mouse so I can rock a little, rock a lot, or stop it totally when I want to study it.



This model was generated using Blender -- a process that I found to have a huge learning curve. I'll post a separate technical thread about that.

--Rik

[WojTek]

Hello Rik,
I've had a look at the 3D model in Sketchfab.
In my opinion, the surface structure comes out very well.
Best, ADi

[rjlittlefield]

Hello Rik,
I've had a look at the 3D model in Sketchfab.
In my opinion, the surface structure comes out very well.
Best, ADi

Excellent!

In the meantime, I think I figured out how to do geometrically correct scaling based on the object's physical size and stack depth. That info is now in the recipe at viewtopic.php?f=8&t=47171 . It's just a matter of setting the object Scale to match the stack's physical dimensions (aspect ratios), and setting the displacement transform Strength = 2.

Assuming that I've done everything right, the physical object has a lot deeper texture than what's shown above. I have added a new Sketchfab model to show the difference. That is at https://sketchfab.com/3d-models/microst ... 97cf0c617b , "Microstructure of a fractured MLCC die, Take #2"



I have also made a model of the whole area seen by the 20X objective. That's at https://sketchfab.com/3d-models/broad-s ... 3575b31903 , "Broad structure of a fractured MLCC die, Take #1". Here is an interesting viewpoint:




--Rik

[WojTek]

Hello Rik,

In the case of the surface structure, the 3D models provide additional information,
which can be observed very well on your models.

The interactive version is not only particularly useful for recognising the structure
of the surface but also the overall layout.

Very well done, I like them all very much.

Best, ADi