Determining carbon coating thickness

[Probeman]

How does your lab determine the carbon coating thickness *inside* your carbon evaporator?

Note: you need to login in order to vote and then you can see the poll results!

[Karsten Goemann]

On our Ladd bell jar carbon rod evaporator we have always determined thickness by measuring the resistance across a clean glass slide using a megohm meter. We also have a "hamster wheel" style sample holder to make sure all samples and the thickness monitor glass slide are at the same distance from the sample.
The system had been "calibrated" a long time ago (before my arrival), probably using brass interference colours and/or carefully weighing the coating.

We have recently installed a HHV Auto306 bell jar coater which has a quartz thickness monitor, but if the sensor is mounted at the same distance as the samples (about 10cm) it looks like it is not precise enough for thin carbon coatings, as there is not much mass deposited. So we may have to move it much closer to the source and that may pose shading issues for the samples... Something we still need to experiment with.
So in the meantime we've built a similar megohm resistance measurement setup for this coater as well.

In the process of doing this I thought I should check the "calibration" of that type of measurement again. We didn't have brass disks handy, so we put thick Au coats on aluminium disks and used those instead. Apparently that should produce similar interference colours to brass, see e.g. here:
https://www.leica-microsystems.com/science-lab/carbon-thickness-evaluation-in-electron-microscopy/
(scroll down for photo with coloured mounts)

As it turns out it takes us 3 of our normal coatings which we thought were 20nm (up to 2 megohms resistance) to get to a coppery colour, 4 coatings for purple, and 5 coatings for blue. This could mean that our coatings are much thinner than we always thought, unless our Au-coated Al behaves differently to brass. I now have some brass disks which I still need to polish to test this.

I'm not too worried about the absolute thickness of the coating as the resistance method definitely produces very reproducible coatings. I guess thinner would generally be better as it seems to provide enough conductivity for our work.

I would be really interested in any comments on the comparison of the different methods, or additional methods if there are any. I guess we could try to do a thin film experiment on the microprobe, e.g. coating something that's heavily absorbed such as boron nitride. Or maybe mask part of the sample and use a surface profiler or AFM to measure across the edge? It might be difficult to get a sharp enough edge for that.

Cheers,
Karsten

[Doug_Meier]

Karsten,

Regarding your masking and profilometry thickness calibration idea, it has been established for Auger and XPS for some time under the name, "Mesh-Replica Method," and is used for sputter depth calibration, though I had used it for coating thickness calibration as well.  More details can be found in the TR cited below.  From personal experience, I can confirm that it works quite well, provided one has an AFM/depth profiler and an XPS or Auger microprobe handy.

ISO/TR 22335: 2007 Surface Chemical Analysis--Depth Profiling--Measurement of Sputtering Rate: Mesh-Replica Method Using a Mechanical Stylus Profilometer

As to the original topic of the thread, another coating thickness estimation (which I picked up from Jeff Davis) is a geometrical method positing that, if one knows what length of carbon rod of a known diameter was evaporated and what the distance to the sample was, one can estimate the coating thickness by calculating the thickness of a spherical shell of amorphous carbon having the same mass as the evaporated rod.  This estimate is usually pretty good, provided shadowing effects don't introduce errors and the evaporator doesn't throw an inordinate amount of sparks.

Doug

[Probeman]

As to the original topic of the thread, another coating thickness estimation (which I picked up from Jeff Davis) is a geometrical method positing that, if one knows what length of carbon rod of a known diameter was evaporated and what the distance to the sample was, one can estimate the coating thickness by calculating the thickness of a spherical shell of amorphous carbon having the same mass as the evaporated rod.  This estimate is usually pretty good, provided shadowing effects don't introduce errors and the evaporator doesn't throw an inordinate amount of sparks.

Hi Doug,
In case you (or anyone else) didn't know, this method is implemented in the little Coat.exe utility that ships with CalcZAF:

https://probesoftware.com/smf/index.php?topic=331.msg1712#msg1712

john

[Probeman]

As it turns out it takes us 3 of our normal coatings which we thought were 20nm (up to 2 megohms resistance) to get to a coppery colour, 4 coatings for purple, and 5 coatings for blue. This could mean that our coatings are much thinner than we always thought, unless our Au-coated Al behaves differently to brass. I now have some brass disks which I still need to polish to test this.

I'm not too worried about the absolute thickness of the coating as the resistance method definitely produces very reproducible coatings. I guess thinner would generally be better as it seems to provide enough conductivity for our work.

Hi Karsten,
Maybe this isn't an issue for carbon thread evaporators, but I've seen the carbon coat conductivity vary quite a bit on our carbon rod coater depending on the amount of heat applied to the rods.  Even when the final color on polished brass is the same.

What I've seen is that if the rods are just brought to a white heat, the resulting carbon coating is almost non-conductive. But if we apply a little extra heat, so that we see some small (several ~cm long) "sparks" from the rod, the coating ends up being very conductive.

These observations makes me suspect that the internal conductivity of the carbon coating is a least somewhat independent of its thickness. Maybe depending on the degree of vacuum, oxygen partial pressure and possibly rod temperature.

Now I don't know exactly what is going on with what I've seen in our evaporator, but it surprises me that simply by monitoring the carbon coat conductivity, you are able to get reproducible thicknesses.

[DavidAdams]

I think this is a discussion that the community needs to take more seriously.

Karstin, when I was at UWA I attempted to determine coating thickness with AFM, but as you pointed out, getting a boundary that is both sharp and not subject to shading between a coated area of a sample and a non-coated area of a sample is extremely difficult. I wasn't able to do it in the time that I was working there.

I think the larger issue, which I have never actually heard discussed before is that the colored-brass method is extremely subjective (it was done quite a long time ago with very large margins of error) and depends on the interpretation of the color from the analyst. Some people, including me and and estimated 8% of males with northern European ancestry, have varying degrees of red-green colorblindness rendering the colored-brass method essentially useless.

I would love to see us in the community working together and perhaps with coater vendors to come up with an accurate, repeatable, quantifiable method to determine carbon and metal coating thicknesses instead of everyone always falling back on the colored-brass method.

[Karsten Goemann]

Thanks everyone for the comments. This is turning into a great thread.

Doug, I agree AFM or even stylus type depth profilers should work well. I used these in the past, the latter for SIMS depth profiling studies. But like Dave I'm wondering how to produce a sharp enough edge for an accurate measurement, as we're talking nanometres. I'll have to purchase that ISO TR you've mentioned. From the public summary it sounds like they're basically using a 75 mesh TEM grid for masking? Another issue is that we have none of these techniques here and mailing the samples around is probably not a good idea.

The geometrical method sounds good as well, at least to "calibrate" another method like our resistance method. Due to the design of our carbon source I can't really measure a length (more below). It may be possible to weigh the carbon rod before and after, but I'm a little bit conscious about mechanical material loss from the rods where they are clamped.

John, yes we're obviously using the conductivity/resistance as a thickness proxy. It seems to give reproducible coating thicknesses, at least when we put the coated monitoring glass slides of different coatings next to each other on a white sheet of paper. Quite subtle changes in thickness seem to be visible to the eye as a slightly different shade of grey. I wonder if it is even possible to measure this as a transmission/absorption and deduce thickness from that. I agree that chamber vacuum, residual gas composition, rod surface contamination... might very well have a significant effect on conductivity. We use the liquid nitrogen trap in our Ladd coater and wait for the pressure to go below 1E-5 torr. Our Auto306 now has a turbo/scroll pumping system and glow discharge unit in addition to the liquid nitrogen trap, so hopefully this will improve the cleanliness.

We also have a different carbon rod source design. We're not using sharpened rods, we're using a single 1.5 mm diameter rod which is clamped at both ends (exposed rod length in between around 25 mm). We crank it up very quickly to almost maximum current (off the chart of the built-in ampere meter that maxes out at 50A) until we start the conductivity starting to climb. The rods mainly seem to evaporate in the center between the electrodes, as they get thinner and rougher there and eventually break in that spot. There is absolutely no sparking. The first coating for a new carbon rod is different - for a similar shade of grey it has higher resistance/lower conductivity. Maybe that's because the new rods are fairly smooth and need to "roughen up" first (it also takes longer to get evaporation started), or there is some surface contamination etc. So we don't use the first coating for critical things. But the following coatings seem very reproducible. It usually lasts for 4-5 coatings and then breaks.

This design was already in place when I got here. I think it is used in at least a couple of other facilities in Australia. I've used sharpened rods systems in the past and also tried it again now with various tip shapes on the Auto306 as it came with it. It seems a lot like a dark art to me to get the tip shapes consistently right for reproducible coatings. I didn't realise you're actually looking for sparks. I always thought these were a sign of uncontrolled conditions, bigger chunks of carbon flying off.

Generally the quartz resonators seem to work fine, but there seems to be the accuracy/precision issue for thin coatings of a low mass element like carbon. Which is why I think I'll need to move the quartz much closer to the source than the samples to get more deposition on the quartz (and have a tooling factor account for the difference in distance). Another problem that is mentioned in the Leica info on the web (the link in the previous post where we've found the C on Au interference colours) is that the heat and light during evaporation seem to affect (increase) the quartz resonance frequency. Essentially the quartz will only give a reliable reading once stabilised after evaporation. With their carbon thread coater where they seem to achieve +/-0.5nm accuracy they're doing pulse/flash evaporation, so they can let the quartz stabilise after each pulse/flash and monitor the carbon build-up that way. That's obviously difficult for a continuous carbon rod evaporation.

So I might continue use the resistance method to monitor thickness buildup during coating, but have the quartz resonator in the Auto306 as a "second opinion" to cross check after evaporation (hopefully track conductivity issues that way) and maybe as least initially also have a gold coated or brass disk as another opinion...

There is an extract from an old MSA list discussion here as well on how the gold/brass interference colours were established:

https://www.bio.umass.edu/microscopy/CarbonFilmThicknessMethods.pdf

This states that the read-blue change at 24+/-0.5 nm was originally determined by Balzers (now Leica) using a "multibeam interference technique for calibration".

Cheers,
Karsten

[Probeman]

Thanks everyone for the comments. This is turning into a great thread.

Doug, I agree AFM or even stylus type depth profilers should work well. I used these in the past, the latter for SIMS depth profiling studies. But like Dave I'm wondering how to produce a sharp enough edge for an accurate measurement, as we're talking nanometres. I'll have to purchase that ISO TR you've mentioned. From the public summary it sounds like they're basically using a 75 mesh TEM grid for masking? Another issue is that we have none of these techniques here and mailing the samples around is probably not a good idea.

Indeed. An interesting thread.  But we need more votes in the poll at the top!

My interest lies more in how to determine a carbon coat "in situ" during the carbon deposition. I guess I should ask Karsten: are you able to measure the sample carbon coating resistance inside the bell jar?  If so, what contact method do you use?  Obviously this won't work for conductive specimens!

My thinking is that if I am attempting to match the carbon coat thickness already on my standards, I'd like to know exactly when the thickness on my unknowns has reached that target, whatever it might be.  In one shot as they might say! Or do you always carbon coat your unknowns and standard together?

[Karsten Goemann]

are you able to measure the sample carbon coating resistance inside the bell jar? ...

Yes, of course. That's the whole point. I crank up the current and monitor the resistance, so I can clearly see the onset and speed of evaporation with the resistance going down. I stop the evaporation once a certain resistance/conductivity is achieved.

As I said we're not measuring the resistance across a sample. We're using a clean, clear glass slide that is mounted at a similar distance to the rod as the samples. That way we have a smooth surface and well defined width and length of the coated area for the measurement and it doesn't matter if the samples are conductive. And we have another indication for the thickness by the (grey) colour of carbon on the glass. We polish that coating off afterward and reuse the slides. They have a bit of silver paint at the ends for contacting the leads of the megohm meter which go through high vacuum leadthroughs.

If I have time I can post some photos of the setup later.

It would be impossible for us to repolish and recoat our standards on a daily basis. We've got >100 standards permanently in the probe. But, again, I think the method is quite reproducible and there's no need to coat samples and standards at the same time. But I absolutely agree that it is crucial to have the same coating thickness on both. Which for me is the whole point of all of this. Have the most robust and easy to use system to produce coatings with consistent thickness, ideally as clean and conductive as possible, adhering well to the samples...

[Probeman]

Hi Karsten,
This sounds pretty interesting, especially for someone like Dave that is color blind.

So you have to replace the glass slide every time, that makes sense. How do you polish off the previous carbon coat but preserve the silver paint for attaching the electrodes?

I'd love to see a picture, but it would also be interesting to see if varying any deposition parameters affects the conductivity, by coating both a glass slide and a polished piece of brass at the same time.

How did you originally calibrate the measured resistivity as a function of actual carbon coat thickness?

[Karsten Goemann]

Hi John,

We normally just use Wenol to polish off the coating on the slide, but something like diamond paste should work as well. The Ag paint wears off over time so we just repaint it every now and then.

That's exactly what I'm trying to do - using more than 1 method to verify thickness, conductivity etc. When I discovered the issues with the quartz sensor (low mass, heat, light) I started using Au-plated Al disks (as originally I didn't have polished brass) and then I realised that the coatings I did on our old Ladd coater might always have been thinner than I thought (see original post).

As I said this system was already in place when I came here in 2006. Probably much longer. I think the Ladd coater is vintage 1977. I think they originally established it by using the brass colour change and/or weighing the rod or the coatings.

[Probeman]

Hi Karsten,
Awesome.

Please let us know when you've done some resistivity vs. absolute thickness measurements. I would very interested to see what you see, especially with respect to different vacuum (oxygen partial pressure) conditions.
john

[Karsten Goemann]

No worries. Hopefully in a month or so we should have the Auto306 set up with our custom sample holders, resistance meter, optimised quartz sensor position etc. That unit should in principle allow us to automate the procedure a bit more, e.g. automatically plasma clean the samples at a certain vacuum, start the coating at a certain pressure...  With the turbo/scroll pump combo and liquid nitrogen trap filled we should be able to operate below 1E-4 Pa.
Maybe have the coating automatically turn off at a given resistance.
Ideally we'd even run the coating at always the same current, if the single 1.5mm carbon rod system is reproducible enough for that. We'll see.

[Karsten Goemann]

Here is a photo I've taken with my phone of a clean resistance measurement slide and a few coated ones. It's difficult to take a good photo due to the reflections. I've also tried putting them on a flat bed scanner (second image) but that's not much better.
The slide on the left is after re-polishing and before coating.
The slide on the right was coated using a brand new carbon rod. In spite of being darkest (thickest) it shows the highest resistance/lowest conductivity. We know that and use the first coating only for non-critical samples.
The 3 slides in the middle from coatings 2, 3 and so forth and all show subtle variations which are probably mainly "operator tolerance", i.e. depends when exactly the person doing the coating turns down the current. As we run at a very high current this is a short time window (probably less than a second). Which is why it would be nice to automate this. But the measured resistance corresponds to the coating thickness - the subtly darker slide (second from the left) has the lowest resistance etc. In absolute terms these variations probably are negligible as far as the analysis goes.

[Karsten Goemann]

Here are a couple of photos of the resistance measurement slide in the sample holder of our old Ladd coater.
You can see that the sample holder is due for a clean...