Atomic Layer Deposition of copper - If you like sputtering, you'll love this!

today on applied science i'd like to tell you about a technique called atomic layer deposition this is a really cool technique to build up a layer on something like a microscope slide or a piece of glass and we're going to build it up atom by atom and we control this by switching back and forth between two different gas chemicals that we flow across this glass substrate so you may have heard of other techniques like sputtering or physical vapor deposition this is another one in those bag of tricks of making a coating on something but atomic layer deposition has a lot of cool features and some drawbacks as we'll see so let's open up the furnace here and see what we got this is the first time i've looked in here that's been running overnight one of the um features of atomic layer deposition is that it's very slow so this has been running overnight all right looks like we've got some decent stuff here i'm very happy this is the first time i've seen it and i am going to touch it with my hands because this thing is all over but let's just you know see what we've got here nice part of the reason that i make these youtube videos is to experience the highs and lows of taking an academic reference and trying to put it into practice so this is not even you know a fifth of the failed attempts that i've got and after working on this thing for you know a few weeks on and off we're over here and i'm really happy with this i think this looks really good it's by far the best one i've made which is convenient because i was going to make this video anyway this morning regardless of what came out of the kiln so you get to experience the happiness the highs that this thing's actually working however check this out i just told you that this technique is amazing because it builds up a very uniform layer atom by atom and if you search for you know atomic layer deposition on wikipedia it'll say oh yeah we lay down a layer of atoms and then we lay down another one and so it's perfectly uniform and one of the things that ald is used for is coating very deep holes so imagine you had a hole that was you know one millimeter in diameter and a meter deep ald is actually able to coat the inside of that hole completely uniformly with a metal or ceramic or whatever you're doing with it but as you can see you don't need expensive lab equipment to tell that this is not evenly coated there's way more copper on the ends here than in the middle and similarly on the glass bottle it's a little bit better it's a little bit more uniform but you can clearly see through the glass at the neck of the bottle and even down here it's it's definitely not uniform so what's the deal like i say you know taking academic papers and trying to make them work in the real world is kind of what's fun about this and figuring out where the imperfections are so clearly there is some technique involved the system doesn't work perfectly all by itself and in the span of you know coming up with all this i'm going to relate to you all the stuff that i found and some of them are pretty interesting problems some of them pretty dumb problems too and so we'll get to that one first and get that one over with and then you know we'll take a look at some nice looking copper coated stuff let's take a quick look at some of the other techniques to make a coating on something and then you'll see why ald was such an exciting technique you know if it works so in the beginning there is evaporation and the way this works is you just take a material and heat it up in a vacuum and eventually as far as i know all materials vaporize when you get them hot enough and the atoms or molecules fly out of your hot pot of stuff here and go in a straight line and hit the substrate the thing that you want to coat and freeze there they condense and become a coating on top so this is very straightforward it has to be done in very very high vacuum if you had any gas molecules between your source and your substrate they would get in the way and cool down these hot molecules so that wouldn't work like if the molecule gets cold on the way here it won't stick to the surface because it's already condensed and made sort of a metal cloud as opposed to a metal layer on your thing so you need really high vacuum and another upshot of this is that there is no mechanism for these hot atoms and molecules to change direction so once they leave the pot they're going to go in a perfectly straight line until they hit something cold this could be good like for example if you put a mask in front of this substrate you can have very nice sharp details the shadow will be perfect but if you're trying to make a uniform coating it's not so good or if there's microscopic scratches on your substrate even the scratch is kind of like a crevice i mean at the microscopic level and no matter how you shine your beam of of um you know atoms or molecules in here the crevice will actually shadow itself and you won't end up with a perfectly uniform coating so if you're trying to make an electrode or something you really need this thing to be completely coated so it doesn't short out or if you're making microchips or something you want a perfectly even coating so partially to address this and also other things the technique of sputtering is often used so instead of operating in vacuum we operate at lowish pressure say 50 militor let's just say for example and the way sputtering works is we use an inert gas like argon and accelerate those argon first we ionize it and then accelerate those argon ions into the target which rips off a little piece of it let's say it's copper metal or titanium or something and the cloud of stuff above this sputtering area is very hot it's plasma so even though there's tons of air of argon atoms and or you know moving around in here they don't actually cause the target molecules to cool off on their way up to the substrate so we can get away with having these things bump into other gas molecules on their way up because this whole thing is very hot it's a plasma and the fact that we're operating at 50 militor means that you know the molecules have time to change direction and it's not quite as directional as this one okay you can also kick things up a notch for example instead of putting argon in or in addition to argon we could add something like oxygen so if we were sputtering titanium and we had argon and oxygen up here when a titanium atom gets ripped off the surface it can react with the oxygen and become titanium dioxide then when it hits the surface we've now formed a layer of titanium dioxide and that's much easier than trying to sputter titanium dioxide itself but the point is that you can have a chemical reaction in this cloud of plasma and you know it opens up possibilities to make a whole kind a whole bunch of different compounds here by adding a process gas so kind of extending that idea this chemical vapor deposition became a thing where you can add two different gases to your chamber and this operates at higher pressure again let's say 10 torr and the chemical reaction here has to be set up such that your substrate is hot and the two gases are reactive enough and you put them in at the right quantities and the right pressures and the right flow rates and everything else you can control this process so for example you could put in you know oxygen and something else they combine and form an oxide coating on your substrate the neat thing about this is that since we're at higher pressure again the uniformity is very high because the gas flows all around it so if evaporation is super directional sputtering is yeah pretty directional cvd is not directional at all because the gas can flow around everything but there is still some amount of control not happening here like for example your substrate might be very slightly hotter in one region or there could be a surface imperfection that causes the reaction to happen faster or slower so even this is not ultra ultra uniform and that's where ald comes in in this process we start with just one precursor and it coats the entire object and we can take our time like we can spend a few seconds coating the substrate with the precursor a let's call it component a then we pump all that out of the chamber and then we add component b which reacts with all of this a component that's stuck to the surface and makes the product that we want then we pump all that out and go back to a which lays down a fresh layer of that and we do a cycle and cycle so every time you go through an a b cycle you end up with one more atomic layer of whatever we want so you can see the difference here if there were a really deep hole like a blind hole in the substrate with cvd what might happen is these process gases get in there and they get used up at like the top of the hole so they never really get down to the bottom of the blind hole because they're reacting with each other and it's depositing the material kind of as quickly as it can get to the surface but with ald since each gas only forms a monolayer then we can take as much time as we need for the gas to diffuse down to a deep hole and it's just truly perfectly uniform at least in theory it is so the next question is why does this a component not form like a super thick layer like why does it only form a monolayer if we're adding more and more a gas to this chamber and we're at 10 torr and the substrate's hot why doesn't it just build up a huge chunk of stuff on there that is because the substrate is actually hotter than our source so the mechanism by which we get this monolayer stuck on our substrate is not condensation because this is actually hotter than where this source gas a is coming from so it's actually just just plain old adhesion like we basically have to carefully clean the substrate and choose our process gases so that it sticks there in just one layer and once one layer is down the the gas a is not sort of attracted to itself once you have one layer stuck to the surface it's no longer sticky and it doesn't grab any more due to the temperatures and chemical affinities involved so you do have to pick the chemicals carefully this doesn't work with just any random set of chemicals and choosing chemistry for ald is of course a huge topic of research so let's see what this thing looks like in real life on the left we've got argon coming in the yellow hose and pure hydrogen coming in the red hose and we actually have three flow controllers that the computer is controlling via a teensy so the teensy sends serial commands to those flow controllers the alley cat and the flow controllers regulate a and b we turn one on to get a in and then we turn b on to get the other the reason that we have three flow controllers is because we also have to have a sweep gas that sweeps the chamber of all the a before we switch to b we can't have a and b in there at the same time and there's no other mechanism that we can use to quickly get the chamber free it's true we could use a vacuum pump to suck it all out but that would take a very long time to get all the way back down to hard vacuum and then add a and then go all the way back to vacuum and add b it's actually much quicker to just have a constant flow of argon which doesn't participate in the chemical reaction so that's why there's three there and then we feed them into this quartz tube which is in a furnace and interestingly the furnace is a little bit colder here which is where we're evaporating the copper one chloride and we use a stream of argon to sort of pick up the copper chloride vapor and it's hotter in the middle of the furnace which is where the deposition is actually happening so the evaporation is happening at about 350 c and the deposition is happening about 415. so clearly we're not don't think of this as like we're boiling something off over here and then condensing it over here that's not what's happening it's actually this just this surface adhesion thing and i have temperature control or temperature monitors to sense the temperature in the middle of the kiln and i even went as far as putting a thermocouple feed through through this spark plug to measure the temperature right at the evaporation source we've also got a pressure gauge here so we can read off the pressure and then on the other end of the tube we've got a vacuum pump and a needle valve and the idea there is that we can set up a flow rate and then close the needle valve slowly until we achieve the pressure that we want so it seems like a very simple setup and it is but that's because i've already debugged the problems and a boyo i had a ton of funny bugs to get through this we'll start on the left here with these flow controllers you do need flow controllers to do this because remember we're building up only one atom per cycle so those copper articles that you saw at the beginning of the video were about 800 complete cycles and each cycle takes 30 seconds so there's really no way you can manually do this with valves you need a programmable flow controller to switch back and forth between the gas flows and these are actually pretty expensive and difficult to get so new these are about a thousand dollars a piece or even kind of used used good ones with a nice display like this for about a thousand dollars i got a killer deal on these on ebay and it took about a week or two to figure out how to use them so nicely enough alicat actually provides their programming manual online but these are so old that the protocol was not operating quite the way they said it was and i had to do a lot of reverse engineering and figuring stuff out and anyway it took a while but eventually we got these running and i verified the flow rate by bubbling some gas through a tube also confounding these valves leak a tiny bit i can tell because if i unhook from the gas supply and put a cap on there and then pump down the chamber the pressure is quite low it's good but as soon as you either uncap this or put a slightly pressurized gas in you know the pressure starts rising so that's just something i had to live with or maybe i'll try to fix it someday i really like these taper or these flare fittings these are really quick and easy to deal with mcmaster sells hoses and fittings and you can really quickly put together a system by having those flare fittings the next problem i was using the wrong chemical so the way this works is we've got to get this copper chloride vapor so i'll pull this out so you can see the setup we want to basically turn on and off the stream of copper chloride vapor very quickly so one way you could achieve this is to have like a hot plate inside there and we raise the temperature to get copper chloride vapor and then we let it cool down that would work but it's just way too slow like we really need to turn the vapor stream on for five or ten seconds and then turn it off and have a very definite on off sort of cycle to that so the way that we can achieve this is to have a flow of argon through a test tube that's inside there kind of like this and one end of the test tube is connected to our pulse gas which is argon and inside the test tube we have our chemical and on the end of the test tube there's a little cut it's just a hole i cut in there with a diamond bit so the idea is that when the gas flow is on we get copper chloride vapor coming out of here because it's hot so this thing will be filled with the vapor and when the flow is zero or close to zero nothing comes out or goes in because that's another problem so so when i first read the paper weeks ago it says oh yeah use copper chloride okay so i rushed over to the shelf and i was like ah perfect i've even got it in stock copper chloride right yeah here we go and then you know i know the difference between copper one chloride and copper ii chloride but there were a few confounding factors that kind of led my mind down this path where i wasn't thinking in the paper they mentioned our sample was contaminated and it was bright green so then we washed it in acid and dried it over the weekend and it turned gray and i was like okay mine's green too no biggie so i also washed it and put it on the hot plate and mine turned brown and i thought well you know it's i mean i bind the stuff off ebay or amazon or something it's probably a teeny amount of contamination sometimes just a very small amount of contaminant will cause it to be colored and i thought yeah it's just brown coloring in there no no big deal so for so for a long time i was using copper ii chloride also known as cupric chloride when in reality this whole process doesn't really work unless you're using copper one chloride oh i see yeah it's a different chemical so you know when you make a dumb mistake like this if you're like me you try to fix it faster like the dumber the mistake the quicker you try to recover from it whereas if it's a really kind of honest mistake you feel like well hey i did my best but when it's dumb like this you try to get in there and fix it fast so i i ordered the right stuff off ebay but in the meantime i really wanted to get this working so i searched for a synthesis of how to make copper one chloride as it turns out it's pretty easy you just mix some of the copper ii chloride in water and you mix that with some sodium sulfite in water and it forms this really nice precipitate very nice fine white crystal powder and you vacuum filter it and you wash it with a little bit more acid and alcohol and you dry it in air and you're done you end up with really nice very fine white crystal copper one chloride another problem i had is that this tube furnace is quite large like in the literature most tube furnaces are maybe 50 or 20 millimeters in diameter and this is like i don't know 40 or 50 or something or more than that even 70. and so in all the flow rates i have to sort of scale them up from the papers but a lot of things don't scale up like that very well and you know it gets to be messy so i use a small test tube like this to get the vapor out into a much larger diameter tube and that's not quite how the paper showed it but there's another challenge here we have to keep the hydrogen away from the hot copper chloride or else we'll just form copper metal in there so the reaction is copper chloride plus hydrogen yields copper metal and hydrogen chloride gas or otherwise known as hydrochloric acid when it meets a water molecule so if we were to have all this stuff interconnected with the hydrogen blowing across everything we would just end up with a little mound of copper metal and that would actually stop the reaction because the copper metal would coat the chloride and then nothing more would evaporate and we'd slow down to nothing so the reason that we introduced the hydrogen in this small tube is to sort of keep it away from the copper chloride so my idea was that the sweep gas is introduced at the end of the tube over here a constant flow of about 500 standard cubic centimeters a minute the pulse gas is low it's about 75 standard cubic centimeter submitted and the hydrogen is introduced like further downstream i don't know if this idea worked super great or not as you can see there is still quite a bit of copper metal forming on the outside a teeny bit on the inside of the copper chloride tube there too but anyway things started working at the end so i think it's basically okay on the exit end of the furnace you can see there's this nice rainbow haze forming on the inside and if i pull the tube out a little bit further you can see there's this definite onset of the hazy area and then it kind of tapers off and we have this really nice interference this is either excess copper chloride that didn't get used in the reaction and so it's it condensed here where the tube is very cold on the output side of the furnace or it could be a contaminant that got evaporated away or maybe it's even some other reaction product or something but it does seem weird that the whole tube is hot obviously it seems like there would be copper it would just kind of use up all the gas so this might be a contaminant or something else i'm not really sure but since i mentioned that the byproduct of this reaction is hydrogen chloride your vacuum pump is going to be sucking down hydrochloric acid and so you're either going to want to put a filter in there or you know change your pump oil or use a pump you don't like very much or something keep that in mind as well if you try this in modern industry i don't think anyone uses copper chloride it's considered the one of the original things that ald was done with and it's good for demoing because it's not toxic most of the chemicals that are used in industry are organic metallic substances if you're going to deposit a metal like tetraethyl lead for example is an organic metallic and that's the same stuff they used to put in gasoline to make it leaded the problem with organic metallic substances is that they're super toxic because they're readily absorbed by the body in a lot of cases and some of these compounds are really complicated they're huge molecules and very uncommon you would not get them unless you were looking for ald precursor supplies so then they're hard to get and expensive and everything and demoing the system with copper is nice because copper is you know really handsome metal and copper chloride is a very easy relatively non-toxic thing to use to demo the process the last really big problem that was surprising to me was the surface prep so since this is a very since this process is so sensitive to the surface like for example if you have an object in there that doesn't absorb copper chloride because of chemicals you know affinities then it won't be coated it's as simple as that so dirty glass may not actually be coated because the dirt doesn't catch enough copper chloride to get this whole reaction started so cleaning the glass is critical and even though i've done a fair bit of critical cleaning for other projects this one required a level much higher than i have needed in the past at first i started with my typical you know basic alkaline you know ultrasonic clean and then water and alcohol and everything that didn't work at all like even though i could tell there were other problems due to the you know the wrong chemicals being used and everything else the coatings were still really just nasty and very inconsistent and everything so i figured okay well i'll break out the big guns we'll go to plasma cleaning i mean this is a great technique because you can do it in situ so you take your samples and clean them in the normal process then load them into the chamber and ignite this plasma i was just using air but oxygen would be even better and the idea is that these energetic oxygen molecules or nitrogen molecules hit the surface and oxidize and blow away the dirt that's there and this is actually a really good technique i'm not sure why it didn't work in my case but in fact it did not i did another run and it was still broken and it's possible that the plasma wasn't energetic enough or it didn't you know it wasn't enough oxygen or whatever but i also thought maybe the problem was that the plasma was getting into the copper chloride vessel so then the next run i did plasma cleaning outside the chamber and put the stuff in and that still didn't work so what ended up working was good old rca clean this is a really aggressive chemical cleaning process where you mix ammonium hydroxide with hydrogen peroxide and it makes this really vicious bubbling bath that dissolves oxide or dissolves hydrocarbon contamination on the surface then you rinse that one and put it into a nasty and even nastier bath of hydrochloric acid and hydrogen peroxide and that dissolves all kinds of ions and you know just touching the tweezers to the top you know instantly dissolves a little bit of metal off the end of the tweezers you know stupid me and so i took some plastic tweezers and got them out of there and this cleaning process was pretty good i still had problems but that was mostly due to the air nozzle that i used to blow these off even though i have an air purifier with drying beads and filtering it's still not clean enough to use for this level of process and don't use those handheld dusters either those things contain a huge amount of contamination of bitterness but even besides the bitterness they intentionally put in there i think there's other oils and hydrocarbons you cannot use those compressed gas dusters to clean something after you've gone through the trouble of doing rca level cleaning it's kind of funny after spending so much time cleaning a surface to this degree when you go back in the house to you know clean your your dishes in your cups and stuff that's not even really cleaning like that's just kind of scraping off huge chunks of food and putting it back away again like it's it's a whole different game of what clean is and for a process like this where you're relying on copper chloride molecules to absorb to the surface it really just has to be incredibly clean like i say this is the most sensitive process that i've ever worked with that requires such a high level of cleaning so the last unsolved mystery that i'm going to leave you with is these glass bottles if you look they all have the same pattern on them the coating worked pretty well from about here to here on every one of these bottles but the neck and the top of the bottle and the very bottom have a different character to them either there's a line there it's not coated as well and all of these were cleaned in different ways some were plasma cleaned some were wet clean some were etched some weren't some were coated with the wrong chemical copper ii chloride some were copper one chloride but they all had this weird pattern in there even despite very very aggressive cleaning and one of these i think i did an acid piranha sulfuric acid piranha which i didn't talk about but that one i cleaned out in one of them so all these different surface preps still yielded this very weirdness about it and you know i used different flow rates and everything so this was not like i put them in different orientations in the chamber so this i don't think this had anything to do with what i'm doing to the bottles i think there's something inherent about the glass that makes them more likely to accept the copper chloride in this region i have no idea what though it must have something to do with how these bottles are manufactured they all came out of the same lot i have a huge box of these things and i just thought that was very strange but it it also makes me think that this whole ald process is super surface sensitive like it may work on certain surfaces silicon dioxide pure quartz might be fine but certain types of glass may have too many ions in it or the surface just doesn't have enough uniformity to work with this process so again it's like one of these things where you read on wikipedia oh ald is perfect it makes this perfect atomic coating but then when you get down to it it's like well it only works on certain surfaces or it depends on your precursor chemicals or you have to heat it up to a certain temperature and then it changes in the cleaning process and so on and so even though it sounds like a downer it's it's actually great in a way to just learn what the limits are a good analogy is for ees when you learn what an op-amp is it sounds like this amazing thing and then when you get out into the real world you realize wow there's hundreds or even thousands of op-amps in existence because they're all imperfect to some degree and choosing the right one is kind of what engineering is all about that's that's what makes it interesting is working with limitations and figuring out how to get what you want okay well i hope you enjoyed that and if you have any questions about ald now is a good time to put them in the comments since it's all fresh in my mind alright see you next time bye

Channel: Applied Science

Views: 434,109

Keywords:

Id: ih_D6OLzqo4

Length: 27min 1sec (1621 seconds)

Published: Sat Jul 31 2021