Applied Science - 2021-08-01
An explanation and demo of atomic layer deposition (ALD) of copper metal on glass. Precursors are copper(I) chloride and hydrogen, processed in a hot-wall tube quartz tube furnace. 10 torr operating pressure 500 sccm argon sweep/purge gas constantly flowing 75 sccm CuCl argon pulse gas (17 seconds including flow controller lag) 100 sccm H pulse gas (14 seconds including flow controller lag) 7 second purge time between pulses 100mm quartz tube furnace diameter 415*C deposition temperature 350*C CuCl evaporation temperature Substrates are mostly borosilicate glass cleaned with RCA clean The "good" samples shown in the video are about 750 cycles (about 9 hours Main ref: https://sci-hub.se/https://doi.org/10.1002/cvde.19970030107 Also helpful: https://sci-hub.se/https://doi.org/10.1063/1.5087759 https://sci-hub.se/10.1149/2.0261501jss https://sci-hub.se/https://doi.org/10.1016/0022-0248(92)90191-K Alicat flow controller manuals (hard to find via website navigation): https://documents.alicat.com/manuals/DOC-MANUAL-MC.pdf https://documents.alicat.com/manuals/old/Gas_Flow_Controller_Manual_rev7.pdf https://documents.alicat.com/Alicat-Serial-Primer.pdf RCA clean: https://en.wikipedia.org/wiki/RCA_clean CuCl synthesis: http://wwwchem.uwimona.edu.jm/lab_manuals/c1901exp35.html Support Applied Science on Patreon: https://www.patreon.com/AppliedScience
The flow controllers probably use a proportional valve which are notoriously leaky when off. The usual way to deal with that is just to put a regular solenoid valve (which are not leaky) in front.
@Frédéric Boivin I believe there are serially programmable controllers, that would make for an interesting experiment... I think the flaw in this theory is that Argon must flow at all times. So possibly one controller and one regulated supply? Unless different rates of Argon are needed for different gases; then two controllers.
@Frédéric Boivin Most flow controllers are thermal capillary types. The manufacturers sort of pretend that they do a different calibration for different gasses, but they actually just apply a factor (which you can find in the manual). How well this works can depend on the design, and you might have to zero the meter between switching gasses.
These days MEMS type MFCs are getting more common, which don't allow for this trick as easily.
Alicat specifically is a "Delta-P" style MFC, which advertises that it will work well with an incredible amount of gasses, and has a nice interface to allow you to select the one you want.
on the systems I work on, we actually have 2 valves, 1 before, and 1 after the MFC. This enables pumpout between cycles, or to prevent releasing hazgas into a vented chamber, and thus the atmosphere. You dont need TEOS or even worse, Silane getting in your face, or out into the air
@Frédéric Boivin You are correct, most MFC's are calibrated to N2 or AR. To use them with other gasses, you will input a calibration factor to whatever control system you're using, and the MFC will scale properly
😊
Other channels: "Check it out, I made plasma!"
This channel: "I tried two different ways of plasma-cleaning as a prep step for my actual project but neither was sufficient so I had to try something else." *shows 20 second clip of really rad looking plasma flow *
@SJAAKSJAAKSJAAK once we get the energy grid sorted environmentally, it’d be so cool to have plasma based dishwashers! What a light show that would be! Being the restaurant dishwasher wouldn’t be the begrudged kitchen job it currently is 😜😸👍👍
I feel attacked
@Magneticitist haha yo I checked out your channel and it's actually pretty cool too.
@Squirrel 2020 lol thanks
@nitehawk86 Science, nothing. I'm an artist.
Borrowing some tricks for this channel makes me look like a sorcerer. Art is trapped in 1522..and Applied Science is in 2522.
I remember sophomore year of high school, I asked my chemistry teacher if was possible to physically measure the size of a single atom. She told me about copper atomic layer deposition and I found it fascinating. Great video :)
You can test the bottle surface difference hypothesis by trying to coat a broken one, just a (flattish) section in the area of most variability should make it clear if the bottle geometry is relevant, or if as I suspect there is some diffusion of some element into or out of the glass from the mechanical characteristics of the process line where they are made. This could actually be useful, hint at a pretreatment step to enhance the effect, if it is boosts the deposition probability.
I'm a PhD candidate in Materials Science at CU, my thesis is on ALD.
One of the issues that you are also running into is the nucleation delay of ALD metals. It's well known in our lab that if you wanna deposit a metal, it's better to deposit a Al2O3 ALD layer in order to create a small layer that allows a better nucleation. We deposit tungsten, and you need ~20 cycles before you see a film growth.
The other is probably a not-perfect ALD process. As other people mentioned you probably need a very long purge and better dosimeters.
for experience, leaks make very weird ALD films.
Emanuele, stochiometric surface reaction is the only way to ALD, true ?
CVD may deposit amorphous or nanosheet layers if precursor not strict ( as Crashy, above comment)
substrate needs to be hotter (local heat) than gas flow temp, with nucleation pretreatment ( TiCl4 ) only how to guarantee 1 layer for TiCl4 ?
what is best commercial polished substrate (Si ?) Have u found Si nitride isolates phonon / electron / hole puddles that form on Si02 / Si ?
@Rus Talisin oh yeah. If you remove the SiO2 from a wafer you see that H-terminated sites gives islands, not monolayers. Basically, If there are not enough nucleation sites you'll have islands.
Then yes, on ALD you are mostly limited to stechiometry, you can't go far from it.. while CVD you can add basically all. SiO2 with dopands? Adding NH3 to precursor and you have SiON or better SixOyNz..
If you have leaks, like here, you definitely doing CVD =D plus the metal nucleation issues.
I once did experiments on monolayers of Cu phthalocyanine.The surfaces had to be cleaned using boiling sulfuric acid and then triple rinsing with very (very very) pure water. I think it's called Type1. Any other grade of water contained too much contamination for producing a good monolayer after rinsing. I'm not sure how you rinsed after the pirhana, but that might be a possible cause for the pattern in the vials.
Great job! The patterns of deposition that you get on those vials, is indeed due to the way they are manufactured. They start as a glass tube, where the edge is torch heated in a lathe untill it softens and then the neck and lip are formed. The same goes for the bottom. So, the reheating that takes place in these two areas, is changing the chemical composition (i.e. taking away Na atoms) and the surface morphology (i.e. sealing nano-cracks). Submerging the vials for a short time in hydrofluoric acid before the ALD, will most likely give a more uniform deposition! Also, refluxing organic solvent over your substrate will help removing organic contaminants. I hope that helped!
I did my PhD in ALD, designed and built reactors. In terms of aspect ratio, ALD can achieve an aspect ratio of 10^5. In ALD, it is all about timing, temperature and pressure. The reactor needs to have really good seal too depending on what is being deposited. The most interesting thing is that the seals become fragile over time because layers also grow inside the rubber seal when exposed to the precursors.
Fun fact: ALD was invented by the Soviets but they only published it in their own journals in the 60s. The Fins took the technology from then and exported it to the world and got all the credit.
This is great. My father would of got a kick out of this. He was a process engineer who built the first plasma etcher (while at STL in Harlow in the UK), and later a CVD coater for aluminium. Now, I deal with optics where ALD is starting become commercial viable and should offer substantial benefits over the traditional evaporative and sputter coating methods.
The pattern on those glass pieces is most likely due to the amorphous structure being aligned differently when the glass was necked down
@Tim Ebert I was just going to say, perhaps topology rather than surface properties
@Cody Goldhahn i don't think they are blow molded, i think they are made from tubes like this https://www.youtube.com/watch?v=vdDSZUjFmb4 this would explain the ring of different material properties at the top since heating the glass causes the surface to lose flux, the only difference between a vial and a threaded bottle is that instead of just necking down on a flat roller, you use a roller with a impression of a thread and interrupt the material into the groove to make the threaded neck.
@Brandon Marsh I think you're right. The straight sides that behaved well is glass original to the tube they started with, and the top part is contaminated with metal from the forming tools that were rubbed against it while molten. I couldn't tell how the bottom was formed exactly, but I think after parting the hot blob collapsed to close it, it must have been placed against a mold and blown to press tightly against it. So, the straight sides are the only part that was not both melted and worked with tools in that state.
@John Długosz the bottom is usually formed by heating a ring of glass and rapidly pulling them apart to separate, then the bottom which is now conical and very thin is just reheated and surface tension does the rest. A lot of people have suggested they are blow molded but I'm convinced because it is Ben we are talking about he most likely has borosilicate vials which are not suitable for blow molding, borosilicate is almost exclusively lamp-worked except for a couple of very specialist applications.
@John Długosz Wouldn't the tools only be used on the outside of the bottle being made, in that scenario?
UV-ozone might be a good in situ cleaning technique for this if your quartz tube passes 185 nm light. You can flow air or oxygen through while illuminating with 185 and 254 from low pressure mercury lamp--look for versions that say they produce ozone, that's the keyword for lamps with an envelope that passes 185. In particular, it's good for producing a uniform oxygen-terminated surface without any ability to sputter, and it can handle significant amounts of residual hydrocarbons from a half-ass ultrasonic/detergent cleaning process.
26:00 bottles were probably made in a way that cools them unevenly, like rolling on the conveyor belt after forming. This would shift the lattice structure of the glass touching the belt, creating the line. Try annealing the glass in the furnace and cool it down slowly.
Also a comment about mass-flow controllers - they are not used this way on an industrial ALD we have. Our system has one flow controller and electically controlled valves for pulsing the precursors. Flow controller is used to provide a constant flow of a purge gas and nothing else. Ammount of the precursor delivered in each pulse is controlled by the time of the e-valve being open, while carry gas is supplied at constant pressure to the precursor vessel.
Depends on the bottle maker. Some are still hot glob spun moulds then excess cut off prior to release and cooling. The bottle at the beginning does look like a 2 step where after the spun mould, it has the neck heated and rolled like you say and definitly changing the latice. I have not tryed personally, however, wouldn't getting the whole bottle back up to almost molten state then cooling realign them?
Yeah, I agree that this is almost certainly from some kind of rolling press that shapes the top and/or trims the excess after forming.
I suspect that annealing it would help but probably can't mitigate the problem entirely. Not without heating the bottle to the point where it loses it's shape anyways.
I wonder what the results would be if you abrade the surface as well or instead?
As for the glass vials with strange patterns, in my experience, the glass will be formed into those shapes using graphite paddles or cast iron tools coated in carbon. Maybe those particles contaminated the glass in the areas they touched and mixed with the glass while it was soft.
You gave me flashbacks to cleaning my 400mm telescope primaries so I could coat them in my home brew chamber. I spent weeks in circular cleaning -> coating -> cussing -> stripping the Al coating -> start over hell. Fortunately I didn't have to resort to your more extreme cleaning measures, a simple but vigorous CaCO3 + H2O wash followed by water rinse and then original Windex applied to a Kim-wipe and used as a vapor worked in the end. 10 years later the mirrors still look pretty good (dusty sadly).
Fascinating video and great presentation! My masters research degree involved closed-field unbalanced magnetron ion sputter plating (a type of physical vapour deposition) to produce amorphous ternary alloys. The beauty of this technique is that you can control alloy composition by optimising the current applied to each of the metal targets (and various other parameters). I would deposit the alloys onto microscope slides and sometimes single-crystal silicon wafers. As you suggest, I also imagine using pure quartz or single-crystal silicon would result in a more even Cu deposition. Although, the purported advantage of ALD is the uniform coating of complex shapes, which tend not to be so readily available made out of pure quartz.
I am unimaginably jealous of the time and money you are spending on these projects. This is the kind of thing I wish i could do for a living
You can. Do it! I believe in you! 🌈
Do you code in java?
@John Dawson Mostly c++ nowadays, but I do still have some tasks in java
Awesome stuff as always! Very timely, working on a CVD project myself and might steal a few of your ideas! :) (in particular the passive argon flow through the copper chloride evaporator... I'm essentially running the hotplate technique you described and it's quite difficult to control). Also, it kills me that most MFCs require an external controller instead of an integrated display. And that the few that have displays are so exorbitantly expensive :(
If you ever want to add a filter in to one of these systems, I recommend using a 1/4 inch flare filter-drier. They are easy to find since they are used in refrigeration systems to absorb water and other contaminants. And they will also screw directly on to your 1/4 inch flare hoses, they have their limitations but will be useful for a lot of applications.
I wonder if there's something about the chemistry of the glass at the surface that differs between the contrasting levels of coating on those bottles. I know that lab-grade glassware is usually made from torch-heated tube stock which is then shaped as needed, and that the localized heating will strip certain atoms (Sodium, I think) from the glass.
To see if the glass bottle geometry is affecting the coating process, perhaps try cutting a glass bottle in half vertically, then attempt the same process.
In my lab’s sputter and vacuum coating, we use a nitrogen gas gun fitted with a deionizer to finish cleaning our substrate. Make sure you are also baking your substrate at 50C for at least an hour to remove any residual water condensate.
Fantastic video! My thesis is all ALD work (actually running a 16 hour deposition right now lol) and watching someone do it all home-made is pretty cool. Now try a selective process :)
EDIT: Just made it to the end - the surface condition being key can be a pain for a blanket film but you can do some really interesting things by exploiting this for selective deposition.
I really appreciate all the time and money you have contributed to this YouTube video. It is so interesting to see real-world work on high technology concepts. I am still trying to make liquid crystal glass circuit panels. Possibly, some day I will make a LCD displays that looks good. You should now try to make a display that says "Applied Science"!
Incredible content as always! Love hearing you explain technical stuff in your videos and seeing your process
I really enjoyed your analysis, especially where you elaborate on what went wrong and your guess as to why. More of these!
It was cool seeing that mass flow meter in use, I used the same device for my senior design project last year. It was too expensive for our allotted budget and we used a loaner from the company we were working with. So you can imagine my thoughts when I first saw you had 3 sitting in view, in one project. I was happy just figuring out how to light up the display, change the units and run air through it! Although I knew it was capable of a lot more, thanks for showing. Awesome video.
At 10:30, beside the kiln, looks like a pressure sensor similar to one we used as well - one with a little green LED on the front where intensity is dependent on the pressure difference.
Absolutely amazing work! A few comments. I would not call ALD a true atomic layer deposition process but it is close. The name is a bit elusive, in practice islands of material are formed during each cycle.
As others mentioned that is probably a CVD process there. A way to ensure the precursors are not in at the same time is through a manifold with a carrier gas and precursors connected to it and controlled by fast acting valves. The precursors need to be in heated jackets or in bubblers to control the vapour pressure. A pump is required down stream.
When you are within the ALD window you should be able to increase both the precursors' exposure/purge time by, for example, 50% and the growth rate should remain the same.
Regarding cleaning ultrasonics and FNA works well. RCA is excellent though as it also chemically terminates the surface which facilitates the adsorption of many metal-organic precursors.
Happy to provide references if needed.
really cool to see your new cleaning protocol. I think many of your previous projects could immediately benefit from it, because as I remember they were often limited by contaminants on your glass slides
Great work! Found this very interesting. I have a couple of old Brooks 5850E massflow controllers. The step response of these is adjustable. If the soft start is enabled the response is very slow, over 10 sec. If set up for fast response it is more limited by the piping, perhaps a second or so. I do not know if other controllers have similar issues. These have an external terminal to fully close the control valve, but I have not used that feature.
I've been wanting to develop this in my lab for several years. I'm really stoked to see you take it on! I can't convince my coworkers to abandon our thermal ALD to commit it to Cu as they still process PV Si from time to time. A custom system would be perfect for what we need. Thank you for the video Ben. My engineer friend and I are big fans of your work.
I'm curious what you do for work
I'm a chemist/physicist studying photovoltaics and chemical catalysts
@Jason C ah, I wish I had a job like that... I don't have the schooling so I'm stuck atm, I need to get in school so bad
@Jason C i personally have an interest in nanotechnology, I've seen some articles showing alot of promise when mixed with photovoltaics
The first thing that came to mind when seeing the patterns formed on the glass bottles was if the internal stresses of the glass might have some effect on the adhesion of the copper chloride. The neck and bottom are the most worked part of the glass, and as such will likely have higher internal stress. You may be able to see these patterns (if they exist) using a polarimeter or similar device, although I'm not sure how well that would work on the complex and curved structure of the bottles.
I love your channel! I love that experts are weighing in on every video too. Giving you shit and providing insights. This is a show of respect. You are doing great, interesting and ego-less work. I really hope you keep this channel alive for a long time!
Great video! it is also possible to have a reactive evaporation, that is how flexible transparent oxygen/water vapour barrier are made at large scale
This reminds me of a paper I read awhile back that indicated they could make graphene on a copper substrate in a way that sounded like CVD, they claimed that it could easily be varied for different quality graphene and was designed in a way that would allow reel-to-reel manufacture of sheets.
However, I haven't heard of anything coming from it despite it sounding amazing. I wonder if you could try duplicating it because it'd be really interesting for someone to manufacturer uniform sheets of graphene in a home lab.
Awesome video. I work in manufacturing and almost all cutting tools come in coated and uncoated variants. You could spend a lifetime studying different coatings, but its good to broadly know the different application methods
Fantastic video, as usual. Have you considered making and/or using a glovebox for those ultrahighly sensitive processes?
Wow, this is awesome Ben!
Please could you try to deposit materials on to a CD which has had it's data layer removed? I'd love to see whether this method could save optical media, however I think temperature could be an issue for the plastic CD.
Thanks for everything Ben! 😎👍
I'm a couple days late watching this video, but I did ALD precursor development in my UG on copper and aluminum oxide. I feel you on surface prep! We would do piranha and HF to get bare Si for our depositions. Another thing to look out for is that you aren't inducing turbulence in your flow. After I left, the lab ran into this problem with a new ALD furnace that couldn't do ALD due to uneven application of gases. This could be the problem with the bottles.
In regards to valves, I think that new furnace used differential pressures of inert gases so that the precursors didn't actually go through valves switching on and off. Unfortunately, this was after my time and I don't remember what they were called. If I figure out the term or a design, I'll let you know
Wow. Nicely explained about thin layer deposition.
When it's comes to the surface contamination, sometime it's important to use ultra pure water(which has conductivity of 18 MOhm). Plus type of the surface plays a major role. Some coatings happens nicely in quartz compare to silicon glass.
Amazing content Ben! Your videos make my day no matter how bad it may be.
I was wondering if radioactive uranium could be used to sandwich b/w semconductors to make a long lasting nuke battery?
Thanks for explaining the different processes in such an understandable manner! That's really interesting!
The inconsistent deposition on the bottle surface must be an artefact of the glass moulding process - either due to the molten glass being deformed differently in different regions of the mould (stretched in the middle and compressed at the ends) or a different surface finish in different parts of the mould because the middle section was turned or ground cylindrical, and the compound shapes at the ends of the mould were made with a different process. It would be interesting to see the result if you flame-polish or mechanically polish a bottle to make the surface finish more uniform.
Finally a new video ! Thanks a lot for this ! It was inspiring and a pleasure watching it as usual !
Nice video! I've never heard of this technique. I wish you better results in the future!
Btw. that copper compounds used in ALD which you shown in video aren't organometallic compounds. They are chelate complexes - there are metal-oxygen and metal-nitrogen bonds. Organometallic compounds have direct metal-carbon bonds.
I love how you make stuff that's way over my head kinda understandable to the best of my abilities.
Looks like those glass vials have been altered chemically by the forming process for the bottom and neck, maybe something has diffused into the surface from the tooling, likely cast iron or graphite. You might try passivating the glass with phosphoric acid, that has worked for me in reducing the effects of the glass surface on the decay rate of active nitrogen plasma afterglows. A little HF-based glass cleaner might refunctionalise the surface too.
I was thinking this too, i think when the bottom and neck are manufactured some form of die is pressed against the glass with some form of "lubricant" to stop the glass sticking, probably graphite, the center section is possibly made out of some massive mandrel drawn/spun/blow molded section of glass that is then cut to length attached between the top and bottom sections.
Or maybe even more likely is there is an external die used to form the bottom section of the glass, then an internal die to form the middle section and then an external die used to form the neck, and the die and its "lubricants" are what allows it to stick better. who knows...
Definitely on the train for annealing, passivating or a quick HF/slow NaOH etch/clean
I've got a couple test tubes that I had coated with metallic copper a couple years ago by passing copper ions over with carbon monoxide while heating the target test tube with a blow torch. I don't imagine I've anything remotely close to what you have accomplished, but now I'm wanting to revisit the experiments I had performed. I'm also going to have to check my spam folder because I don't recall getting a Patreon notice for anything lately. Your attention to detail is always amazing. Take care!
i think your gas flow is quit turbulent due to the high reactor diameter, you would most likely benefit from making it more laminar by passing it trough some glass wool or something within the reactor tube.
ALD chemist here:
Evap/sputtering are directional, CVD is diffusion limited, ALD is reaction limited. A lot of ALD precursors can be CVD precursors under different operating conditions, or even have some CVD components. Flow rate, system pressure, pulse/purge times all affect the growth rate, con-formality, and surface structures of a film.
Crashy McSplodey - 2021-08-01
You are accidentally still doing CVD. There's huge amount of reactants still in your lines after you close the fow meters and instead of 30s cycle time you probably need 30min. To truly stop reactant flow in those lines you need inert gas barriers on each source. Hydrogen is simple with two solenoid valves and two needle valves. Put two T junctions on the line. First (closer to flow meter) goes trough solenoid and needle valve straight to your vacuum pump. Second (closer to reactor) goes trough solenoid and needle valve to your carrier gas supply. Now when you open both solenoids hydrogen has straight path to pump and leftovers in the line will be purged by reverse carrier gas flow. Close the solenoids during hydrogen pulse.
For copper source it's the same principle. Just imagine that the second T-junction is inside the reactor hot zone and and reactant is just before it. You'll need one extra feed trough into your reactor for the blocking gas. This apparatus is usually made out of quartz but you can probably get away with glass or even stainless steel.
Source: I operated and modified experimental lab-made tubular ALD reactors in uni for many years.
When does the video actually start? - 2021-08-02
@Tech Gorilla So where's your research data? All we got in these comments are various opinions. And yours were until now the least useful ones.
Tony Wilson - 2021-08-03
Great input to the discussion.
On his plasma cleaning I think he need to not simply place the parts in the bottom of the tube. No matter what in that case different areas of the parts get different exposure.
I did work in the auto sector (as a control and automation engineer) and they used plasma cleaning for the reflectors in head & tail lights before coating the parts. I helped build one of those lines for Hella (in Australia). For the final deposition (which used evaporation) they had all the parts on this carriage that was driven the same way an epicyclic gear box operates. The parts were on shafts that were extensions to the planetary gears. So the parts not only spun on the shafts but the shafts rotated around the center of the vacuum chamber. That way they got evenly coated.
I don't think he needs to go to that sort of extreme here but he possibly needs to ensure and even flow of the gases and plasma across and around the parts he's trying to coat.
Robin Binder - 2021-08-16
engineering nerd corversations be like ...
jk i like to read it too even tho i understand maybe half
Emanuele Sortino - 2021-09-13
@P McR as PhD candidate with thesis on ALD, I can tell you that you are 100% correct :)
pasha veres - 2021-10-02
@Tiporari Sorta' like he's constructing a mnemonic memory device using stone knives and bearskins? ;)