 Good morning, welcome to the next lecture in our course on chemical engineering principles of C V D processes. In the last lecture we talked about methods of classifying C V D processes particularly based on the operating pressure. So we talked about atmospheric pressure C V D, low pressure C V D and plasma enhanced C V D. Today we are going to talk about another classification of C V D process that is based upon the film that you are depositing as well as the precursors that you are using to produce the reactive materials that result in the film formation. As I had mentioned briefly in one of the previous lectures metal films can also be deposited using a C V D process. The reason that you want to deposit a metal using C V D is that even though it is not the easiest or most convenient process it again has the best step coverage. Now particularly in microelectronics has device geometries continue to shrink because you are trying to pack more and more circuits into smaller and smaller areas you start having these fairly deep trenches where the aspect ratios are very high and if you are trying to get a metal layer into these the deep features it is very difficult to do with virtually any process other than chemical vapor deposition. So the biggest advantage of C V D in this context in the context of metal deposition is that it has very good what we call throwing power. Throwing power essentially means the ability to coat uniformly over widely varying features of a surface and so metal C V D has been around for quite a while. The most common metal that is deposited using C V D is tungsten I believe I mentioned that in one of the earlier lectures. Now tungsten is a reasonably reactive metal and in the past tungsten C V D has essentially been done using the low pressure C V D process horizontal flow LP C V D but one of the difficulties in the kind of well it was basically hot wall horizontal flow LP C V D that was being used but because tungsten is quite reactive the problem was that it would not only deposit on the substrate but also on the reactor walls and so forth. So the one change that has happened over the years is people are now tending to use more of cold wall type of C V D process for tungsten deposition. In fact the wall temperatures must be kept less than 150 degree centigrade in order to prevent tungsten depositing on the walls and the chamber itself. Now as far as the precursors that are used for tungsten the most popular ones are W CL 6, W CO 6 and W F 6. Now W CL 6 and W CO 6 are solids at room temperature but with high vapor pressure. W F 6 is a liquid at room temperature but tends to vaporize fairly easily once you exceed the room temperature. In fact the most common mode of depositing tungsten is to use a mix of W F 6 and H 2. Actually W F 6 itself can obviously give you tungsten solid plus 3 F 2 but the energy barrier in this case can be quite high. A preferred way of doing this is essentially to use H 2 as a carrier gas which will give you H F as a byproduct. This process is traditionally known as blanket C V D for tungsten. It actually provides the most uniform film of tungsten on a surface. The only, there are couple of problems though with this process. The first one is that the deposition rate can be quite small. So it is not a hugely reactive system and so the film thickness tends to be quite small. So depending on your application if you need thicker tungsten this is probably not the way to go. The other problem is actually H F that is produced as a byproduct. Now this again introduces two problems. One is H F is a kind of a toxic chemical so it needs to be treated before it can be let out into the atmosphere. And the second problem is H F also tends to adsorb quite easily on the substrate surface. So this reaction and deposition of tungsten quickly becomes limited by the rate of desorption of H F from the substrate on which it is adsorbing. And so this again can slow down the reaction as it proceeds as more and more H F is produced and starts occupying the interstitial locations on the substrate. It can actually delay the onset of tungsten film formation. So the alternative to this is to actually use Si H 4 silane along with W F 6 to get again tungsten solid. The advantage of this technique is that depending on whether the reactant mixture is tungsten rich or silane rich you can actually get either tungsten or tungsten silicide depositing as a substrate. So you have a way of actually tweaking the composition of the reacting gases to get the film that you want. If you are looking for a layered deposit where you have tungsten tungsten silicide you can actually tune it fairly easily simply by controlling the flow rate ratio of W F 6 to silane. The silane process has the advantage that it has a lower energy barrier. The deposition rates can be significantly higher. However it does require that instead of trying to deposit the tungsten directly onto the substrate you actually put down a layer on top of the substrate which helps improve the adhesion of the tungsten to the substrate. So these are typically nitrides. For example titanium nitride is frequently used as its underlayer to promote adhesion of tungsten to whichever substrate you are trying to put it down on. For example if you are trying to do particularly silicon, tungsten on silicon the adhesion properties can be quite poor and so this intervening atomic layer helps improve the adhesion considerably. But of course this itself you know how do you put down a nitride layer on the surface. That is another CVD process essentially. So metal CVD is typically preceded by growth of an underlayer which are typically nitrides that bond well to the silicon and also provide a good bonding surface to the tungsten and for that actually many times the chloride precursor is preferred. So for example if you want to put down a titanium nitride underlayer and you want to do it using CVD you would set up a process like TICL4 plus NH3 going to titanium nitride plus HCl plus N2. So this would be the first step in your process to put down the underlayer and then you follow that up with your metal deposition process. Again the downside of most CVD processes is the production of these byproducts that requires some consideration and some treatment before they can be emitted. In fact we will see later on that when you look at advantages versus disadvantages of CVD the number one disadvantage of CVD is that the upfront investment can be quite high. It is not just a matter of building a reactor. The reactor has to have many controls both on the input feed and on the operating conditions of the reactor and on treatment of the exhaust. So in terms of the complexity of the system the treatment facilities that are required as well as the controls that are required over the various processes CVD reactors tend to be quite demanding. So what that means is it adds complexity to the design. There are more things that can break down essentially in a CVD reactor. Complexity equals cost. So anytime we talk about adding complexity what we are really saying is it is going to be more expensive. So the upfront capital investment can be quite high in a CVD reactor and also a CVD reactor cannot be run in any kind of environment. It has to have excellent treating treatment and venting facilities and so you have to design the CVD facility keeping in mind what are going to be the potential byproducts. And here again the difficulty is because it is a very reactive environment you cannot always predict the exact composition of the product species. You could have variations depending on I mean even a small change in your input conditions or your operating conditions can result in a significant change in the chemical equilibrium composition of the byproducts. So you have to be kind of constantly alert to that and make sure that whatever system you come up with for treatment and emission control is robust enough that it can actually handle small variations in the composition of the products that are exiting the reactor. So this kind of becomes even more important when you deal with things like metal CVD which is a very reactive system and you have to use reactive chemicals in order to make it happen. Now silicon CVD is probably among the more benign CVD process. Typically you are using silane which breaks, thermally breaks down to silicon and hydrogen. It is not a huge problem in terms of dealing with the chemistry is involved but as you get to more reactive CVD process you have to think harder about how you do it. Other materials other metal that is frequently deposited using CVD is copper. Copper is a fairly reactive metal also but the problem with putting down a copper layer using CVD is that it is not easy to find inorganic precursors from which copper can be deposited unlike for example tungsten where there are both inorganic as well as organic vapor species that can be used to promote the deposition process. So if you want to use CVD to make a layer of copper you have to use what is known as MO CVD which stands for metal organic CVD. It basically just means that the vapor precursor from which you are depositing the film is organic in nature. And the reason you do this is again because there is no inorganic precursor. So MO CVD in many ways is a lost resort. You only do it if there is no appropriate inorganic species that you can use in your CVD reactor. The reason again is that when you talk about metal organic CVD the term organic implies the process of carbon. Carbon frequently becomes an impurity in the film very hard to keep it out. So unless you are intentionally trying to incorporate carbon into your CVD film as a dopant you want to avoid using carbon containing precursors in your CVD process which is obviously impossible to do when you are doing MO CVD. So that is the primary factor that discourages people from using MO CVD but it does have you know several advantages as well because to prepare the vapor precursor is not too difficult. So the way you would do it is you know you would take the metal you would grind it into fine particulate form. So these would be your metal particles and then you would either soak them in an organic solvent could be ethyl alcohol, methyl alcohol, butanol and essentially let the particles react with the metal and produce the metal organic precursor or you would essentially use a bubbler where the liquid essentially you would have a bed of the particulate material and you would let the liquid essentially flow through the bed and interact with the metals to form the vapors. So when you do this you can take basically any metal plus let us say that you are using methyl alcohol then you can make various compounds. For example DMZ is a compound that is used to make zinc films. This stands for dimethyl zinc. So similarly other popular precursors in MO CVD or TMG which is trimethyl gallium which is obviously used to make gallium films. TEG stands for triethyl gallium, TEA which stands for triethyl aluminum, TBA which is tetra butyl arsine. So in each case the this is the MO CVD reactant or precursor. So these are in vapor form and they are fed into the CVD reactor in order to make the corresponding metal film. Of course what you are really trying to do is take the metal and turn it into vapor form right. I mean that is basically what you are doing. So another way you can think of doing this is simply by taking that metal piece and bombarding it with sufficient energy to directly vaporize the metal. So for example you could take a metal and hit it with high energy ions. You can actually cause localized liquefaction of the metal and you can start evaporating the atoms of the metal so that they enter the gas phase. And this was actually the process that was used to make metal vapors long time ago. Does anybody know the term for this? What do you call it when you actually use energy to vaporize a solid substrate into vapor form? Have you ever heard of a term for that? It is called molecular beam epitaxy. This was the process that was in use before the 1960s. It is actually invented by a professor at Yale University Professor John Fenn who won the Nobel Prize for this. Molecular beam epitaxy simply implies that you are using a high energy beam to knock off solid molecules into a vapor phase. And so this MOCVD is kind of a competitor to MBE because it is essentially achieving the same thing, right. It is entraining the metal molecules in the gas phase. So the MOCVD process was invented in the 1960s and since then it has predominantly replaced MBE for CVD of metals because as you can imagine the energetics involved in this process are quite high and maintaining particularly the purity of the chamber is always a huge issue. And the other problem that happens in this particular case is that it does not, it does not etch the surface uniformly. Essentially you are picking spots on the metal and training them whereas in the case of MOCVD you are pretty much reacting the entire bulk material to make the metal vapors. And so you get a much more uniform feed when you are doing metal organic CVD type of processes. So this MOCVD process is now widely used also to make metal films. Now these processes, again the metal organic is a highly reactive system. So you want to run MOCVD reactor under conditions where there is good control over the reactive nature of the system. In other words you do not want to run it at too higher temperature. You want to run it at an intermediate temperature where the deposition process is kinetically controlled rather than purely controlled by transport phenomena. If you recall the figure that we had drawn earlier relating deposition rates to substrate temperature 1 over T versus growth rate, you recall that the general trend is like this. There is a period, a low temperature period where actually as you increase the temperature there is an exponential increase in the rate of film formation. But then once you raise the temperature beyond a certain value it becomes constant. So this is the classic CVD signature. When you talk about metal organic CVD there is actually a small variation to this which is that at sufficiently high temperatures you can actually start to see a reduction in the rate of film formation. Can anybody speculate why that might be happening? Why would you see a lowering of the rate of deposition? Any ideas? The primary reason is that it is a very reactive system. So what that means is there can also be many parasite reactions that can be going on at the same time. Of course a parasite reaction is one that essentially scavenges the species that should be depositing as a film and instead causes it to form a different product. So for example let us say that you are trying to make gallium arsenide and so what you are looking at here is the system itself is you have let us say TMG plus arsine which is giving you the gallium arsenide material. Some of the parasitic reactions that can actually be two kinds. The parasitic reactions can be such that it keeps the gallium and the arsenic from depositing on the surface. So it keeps them in the vapor phase. So they act like a gettering mechanism which prevents the formation of the film. Alternatively what can happen is instead of forming gallium arsenide as your film you can actually start forming gallium or arsenic as the film. So then it becomes an impurity in the CVD film. So when we look at the growth rate of the deposit the overall thickness may actually be increasing as a function of time but the composition may be changing whereas in the case where the depositing elements are actually removed and kept in the vapor phase you will see a net reduction in the rate of film formation itself. And when something like this happens the and that is one reason why you do not want to operate in this range. Typically MOCVD is operated in an intermediate range where there is still a significant effect of temperature and the conditions are not so reactive that you start forming all of these essentially unwanted or undesirable reactions either in the gas phase or on the surface. And for the same reason just like metal CVD MOCVD is also done under conditions where the walls are essentially kept cold. That is again the downside of using a very reactive system. You get high film formation rates but at the same time your entire reactor is susceptible to formation of particles and so on. MOCVD reactors are run in two configurations usually. So the first is the classic horizontal cross flow type of a reactor but the other configuration that is widely used is the stagnation point flow reactor. Now the reason that people prefer the stagnation flow reactor for MOCVD is that when you have the horizontal flow and you have a cold wall so that this TW is much less than TS. As we have said before you can start setting up these recirculation cells which can introduce turbulence into your system and prevent the formation of a uniform film. A lot of those problems can be overcome essentially by using a stagnation point flow reactor because here the flow is predominantly directed towards the surface. So even if your wall temperatures are much lower than your substrate temperatures the essentially the downward flow of the reactant gases is sufficient to prevent significant recirculation cells forming and taking the material away from the substrate. Of course in stagnation point flow also you can have localized recirculation cells but they are contained to a region that is very adjacent to the substrate. So the effect of these convection rolls are not felt through the entire reactor. They are kind of localized and their influence can therefore be kept to a minimum whereas in the case of the horizontal flow CVD reactor once natural convection effects start to show up it is very difficult to contain it. By the way the other way that turbulence can happen in a CVD reactor is that even if you have good controls over everything you know your input, inflow, your temperature distribution and so on. If the geometry of the flow it has certain sharp features you know corners or turning radii that can sometimes be sufficient to induce turbulence. So it is not only the flow dynamics that result in turbulence it can also simply be the physical design of the CVD reactor that can result in unintended turbulence effects. Now that again the horizontal flow design is much more susceptible to compared to the stagnation flow design because the path from the inlet to the substrate is essentially a straight line in a stagnation flow reactor. So the geometry of the reactor itself does not play a significant role in controlling the deposition on the film. If you are going to use a horizontal flow reactor many of the things that we have talked about earlier in terms of the flow dynamics controlling the velocity profile so that you enter with a uniform or plug flow which then turns into a parabolic flow that becomes very very important to achieve in the case of an MOCVD reactor. In the case of the stagnation point flow reactor the primary thing that you would want to look at is what is happening in the vicinity of the substrate. The thing that we want to ensure is that the residence time of the reacting species adjacent to the substrate is kept uniform along the length of the substrate and so you do have to do some CFD modeling to understand how flow is going to happen in the vicinity of the substrate. So again in terms of analysis and controls the focus in a stagnation point reactor is only around this region. So that reduces the complexity of the design and therefore the complexity of the cost. If you are trying to do this in a horizontal flow reactor then you really have to do a CFD model over the entire reactor and you have to control not only what happens in the vicinity of the substrate but essentially everywhere within the reactor. So complexity goes up, cost goes up. Alright so in terms of metal organic CVD so these are some of the considerations you have to keep in mind given the high reactivity of the precursors. The other thing as I mentioned earlier is carbon. When you are trying to deal with systems that contain a lot of carbon, what you can try and do is either make the film itself sort of forgiving of carbon so that the properties of the film are not unduly affected even if there are some traces of carbon adsorbed into the film. The other thing you can try and do is use sufficient excess of the carrier gas to make sure that the excess carbon is flushed away as soon as possible. Absorption is a time dependent process. So the longer the time that you allow for a vapor species to be adjacent to a solid substrate, the higher the probability that it will get adsorbed. So by ensuring that there is a strong flow maintained, you can essentially keep removing the carbon from the vicinity of the substrate as quickly as possible. So providing a horizontal flow actually helps in that case. So the primary advantage and the reason that people still look at horizontal flow CVD as an option is that the potential for carbon adsorption and in general impurity adsorption is lower in a horizontal flow CVD setting compared to a stagnation flow. The same things that work to your advantage in terms of improving the deposition characteristics work to your disadvantage in terms of worsening the impurity deposition characteristics. Any system that is more efficient in promoting deposition is also going to be more efficient in promoting the deposition of species that you do not want. So there is a fine balance. A lot depends on what is your purity requirement. If for example your film has to be more than 99% pure, then you have to make absolutely sure that you cannot have impurities present. Then you may choose to pay a penalty in terms of some of the other characteristics of the system but ensure that you have minimal impurities captured in your film. On the other hand if your film is something that you sell in the mass market, you do not care so much about whether it is pristine, then you are probably better off using the stagnation flow type of reactor which will ensure that your reaction rates are very high, your deposition rates are very high and you know the only thing it cannot ensure is an impurity free deposit film. So these are some considerations to keep in mind when you are trying to design these various reactors, you know atmospheric pressure, APCVD, LPCVD, PECVD, MCVD, MOCVD, they all have different purposes and there is no one design that fits all. You essentially have to look at each of your applications and see what type of process is best suited for that particular film. And I think it is fair to say that CVD is a process that you invest in only if it is worth it. If you are trying to make a very thick film or you do not care how good the film is, you just want the film to be there and so on, you are probably better off not even looking at CVD as a process. I mean CVD is a process that requires chemical engineers to run essentially. You need chemical engineers to design it, you need chemical engineers to run it, you need chemical engineers to control it and to optimize it and so on. Whereas there are many other coating processes that can be done essentially as a turnkey operation which do not require such a you know good understanding of what is going on inside the system. For example coatings, so we will talk a little bit about CVD of coatings in the remaining time in this lecture, in the next lecture. There are many ways to put down coatings on surfaces. You can spray them on, you can do dip coatings, you can do you know thermal coating processes. There are so many, there is electrophoretic coating, there is electro less coating. All of these methods work perfectly well in terms of making a coating on a substrate. So why would you want to look at CVD if you want to simply have a coating on a surface? What are the advantages of CVD coatings? The primary advantage of course is that you can make very thin films. CVD is an atom by atom or molecule by molecule process. So it is essentially what we term a molecular self assembly process. You are putting the film down a molecule at a time. The molecules find each other and they assemble themselves in such a fashion as to give you a continuous film which certainly offers you the attraction of being able to manipulate the coating also at atomic or molecular level. So if you are very keen on providing particularly a layered deposit, a layered film where there is a variation in composition as a function of sectional depth, then CVD may be the only process that you can use. Because CVD is the only process that can give you that atomic layer definition. So in principle you can make a coating where the first molecular layer let us say is silicon oxide, the second molecular layer to silicon nitride, the third layer is again silicon oxide, the fourth layer is titanium oxide. In principle it is possible to do it and each of these can be of the order of angstroms thick. There is no other process that can really give you that level of control. So if you have an application that requires that level of granularity then you have to use CVD. The other advantage of CVD is that it can pretty much coat any substrate you know no matter what its physical features. Particularly when you have these sharp changes in the geometry of the substrate, when you have deep trenches or when you have bumps and blisters on the surface. These features are typically very very difficult to coat with any process other than CVD. So the more complex the geometry of the substrate, the more attractive CVD becomes because of its step coverage and throwing ability. These are the two terms that are normally used to describe the ability of any coating process to follow the contours of the substrate. The other advantage of CVD is it is a fine grained coating and grain size plays a huge role in determining ultimately the functional properties of the coating and normally a finer grain is hugely advantageous. And here again because CVD is a molecular level manipulation process the grain sizes tend to be also very very small of molecular scale. And so it is very easy to get a fine grained structure using CVD as your deposition process. The other advantage with CVD is that you can minimize impurities because it is a well controlled environment you really again have molecular level control on the composition of the film. So you can very very precisely control the stoichiometry of the deposit. I mean you do not just make Si3 and 4, you can make Six and y and have x and y change continuously as a function of depth. There is no other process that really can give you that kind of ability. So the other advantage is that typically CVD films tend to have good strength, good durability characteristics because you are relying upon interlocking layer to layer. The biggest advantage of a CVD process is that thermodynamically they are favored. Equilibrium thermodynamics favors the formation of the coating. So you do not have to force it to happen. As long as you have the right ingredients in place and you have devised the operating conditions appropriately, CVD will happen organically or naturally and the film will start growing. Now that is always good in terms of you know having confidence in your ability to grow a good film on a surface. Now the disadvantages, as I mentioned before the investment cost to first set up the CVD reactor is more significantly more expensive than virtually any other type of system you can imagine particularly for doing something like a coating on a surface. The second problem CVD films are not particularly well adhered to the substrate and the reason for that is that the typically adhesion is being achieved by chemical interlocking. You are relying upon the species that wind up on the substrate to chemically react with the surface and form a film. Now in principle chemisoption forces are much stronger than physisoption forces. However that only applies when you can guarantee chemisoption to happen. But if you have impurities on the surface or contaminants on the surface or by-products that are getting absorbed on the surface. They can compete with the reactant molecules and prevent this chemical interlocking from taking place and when that happens there is no backup. You know there is no physical mechanism to lock the film to the surface. So absorption is something that you always have to, I mean adhesion is something that you have to always watch out for. The density of the film is also lower in CVD compared to thermal coating process. So typically CVD films need to go through a densification process which is a post-CVD baking or annealing process to improve the quality of the film and get the properties that you are looking for. But the primary problems with CVD are the cost and the complexity. Because there are so many factors involved, I mean it is not enough to just control the flow dynamics. You also have to control the composition of the reacting species. You have to control the temperature distribution in the reactor and you have to control what is happening in the vicinity of the substrate as well as what is happening downstream of the substrate because many of the by-products of CVD processes whether for coatings or for other substrates or toxic in nature or corrosive in nature or have other types of environmental effects that must be prevented from happening. And so in the case of for example many coating processes tend to be liquid based where you essentially expose the substrate to a liquid and let the liquid dry on the surface or they can even be solid, for example powder coating is essentially a solid phase process where you turn the coating materials into very fine aerosols and spray them onto a surface in the form of a very fine powder. In situations like that you know the discharge that you have to deal with is somewhat easier to deal with than the case of CVD where the by-products tend to be primarily vaporous in nature. The downside of that is they are very reactive. So if they do get into the atmosphere you know that they are going to cause carcinogens effects, toxic effects and so on. So it becomes very, very important to treat them at the outlet of the CVD reactor and ensure that they do not have any harmful effects downstream whereas liquid and solid discharges are fairly, I mean there are well established processes to be able to deal with them. You can do it physically, you do not have to do it chemically. So there are pluses and minuses to the use of CVD processes in this application but CVD of coatings is a huge industry. It is a multi-billion dollar industry so there must be a reason why people want to do that. So in the next lecture we will talk specifically about CVD of coatings and what are some of the unique features of these processes. Any questions on what we covered today? Okay, see you tomorrow. See you at the next lecture.