 Good morning and welcome to this lecture in our course on chemical engineering principles of CVD process. As I mentioned in the last lecture, we are going to discuss how CVD relates to nanotechnology. As you know nanomaterials are increasingly found everywhere. They have certain unique properties that are very advantageous whether you are trying to take, they are trying to leverage their reactivity or their surface area or their hardness characteristics or their melting point differences. Each nanoparticle has a certain specific characteristic that makes it very attractive for certain applications. For example nano sensors are used increasingly to study the presence of trace amounts of materials, parts with trillion amounts of materials in various environments starting from the atmosphere to the human body. Nano drug delivery is another emerging application where time release of medicines is possible by encapsulating them in a nano carrier. For chemical engineers nano catalysts have huge potential in terms of being able to drive up the available area for promoting surface reactions by several orders of magnitude. Of course nano devices in general particularly the MEMS and MEMS devices microelectromechanical devices nanoelectromechanical devices which combine the functionalities both the electrical as well as the mechanical functionalities of nano materials are also finding increasing application everywhere. So given the fact that nano materials are in great demand there is a huge amount of interest on how to actually make nano materials. If you look at nano in general it is still what I would call in the science realm rather than technology realm. I mean it just hasn't the production of nano materials has not scaled to such an extent that it has penetrated deep into our everyday life except you know in occasional products like you know cosmetics or topical creams or things like that. So what is holding it up? Well the fact is nano materials are difficult to make. Now if you look at nano technology overall there are really 3 important steps to it. The first is the synthesis slash assembly of nano materials. The second is the characterization with nano materials you constantly have to characterize the product to make sure that it is coming out the way you want it and so the characterization is the other important technology and the third one is manufacturing and processing particularly the processing aspect you know one of the difficulties with nano materials again is the tendency to agglomerate. Even if you make nano materials and at the time of production you can ensure nano dimensionality you let the particles just stay in suspension for a while they stick back together and they can become larger in dimensions. So the post processing of nano materials after they have been synthesized is another important challenge. So fundamental to tell to all this is again modeling unless you can capture the theory of what is happening how are nano materials being formed? How do you characterize their stability over time? How do you characterize their ability to perform certain operations that they are intended for? Unless you can model all this and develop a predictive ability you are really not going to be able to either control the process or optimize the process. You are kind of going to be at the mercy of you know day to day variabilities in the process or in the purity of the materials or in operating conditions and so on. So the challenges are really two fold one is to develop a manufacturing process that is sufficiently scalable we want to be able to make tons of nano materials per day not a few grams for lab use and the second is to be able to model these manufacturing processes with sufficient rigor where we have some predictive ability over what you are going to see at the end of the process. So given these constraints chemical vapour deposition has actually emerged as an attractive tool to make nano materials because it satisfies both these conditions it is scalable and it is simulatable. Now the other approaches to making nano materials essentially well we can classify them in two ways bottom up and top down the top down methodology you basically take larger particles and break them down into finer particles. So you take micron size particles and by crushing or grinding or sonication or whatever method you reduce their size to nano dimensions so that is only one way to go and it has its pluses and minuses. The other way to go is to build up the nano materials from bottom up which means start with atoms or molecules and keep attaching them to each other until you get the particular product that you are looking for. CVD is obviously a bottom up process but there are other techniques particularly the molecular self assembly techniques using which you can essentially make designer product. You have the ability now with certain nano dimensional tools where you can actually manipulate atoms and place them in certain configurations so that you get exactly the structure that you are looking for. These products are also called Langmuir Blodger films or molecular self assembled films and they have this distinct characteristic that you have absolute control over the structure as well as the morphology of the product. So they work great, problem not scalable. I mean this is wonderful for you know making a few milligrams of material for lab use but it is not a product, it is not a method that is going to give you large amounts of nano dimensional material. The next best thing is CVD where you essentially deposit the film on the surface but what you try to do is do not let the film grow to a size larger than a few nanometers. In other words you should have certain checks and balances in your deposition process or certain features on your surface which will ensure that the product once it is deposited does not become a continuous film over you know many millimeters or many even mining micrometers but somehow it is curtailed after it grows to a few nanometers. Now there are couple of ways in which you can do this. One way is to deposit the material using CVD and essentially provide certain energetic barriers on the film which prevents the surface migration from happening. Make it difficult for the atoms to keep jumping from one space to the other and this can be done by crystalline orientation, it can be done by providing gradations of surface energy on the surface. There are different ways you can do this which is essentially focuses on preventing the surface migration or surface diffusion of the deposited molecules to form a continuous film. The other way is to do something in the gas phase. As we have discussed earlier you can turn a CVD reactor into a aerosol reactor by allowing heterogeneous nucleation to take place. So you try to form nano dimensional particles in the gas phase and again freeze the process so that these particles do not keep growing to larger sizes and then direct these nano dimensional particles to deposit on the surface so that you essentially form these discrete nano particles on the surface. So these are methods essentially using CVD or also what is known as CVS chemical vapor synthesis reactors to make nano dimensional products. Now in terms of again modeling and simulation of such reactors to make nano products, this is something we discussed in one of the earlier lectures. The modeling of any CVD system requires multi scale modeling right. There is the so called macroscopic system which then breaks down into the mesoscopic system which then breaks down into the molecular system. When we talk about nano materials you essentially add another layer of complexity because now you have to have a model essentially at atomic level which will capture how the growth is curtailed to a discrete particles rather than a continuous film. So you add one more element to the model. In fact another way to look at what we have discussed in that class is to plot your let us say your length scale to a time scale on the y axis. So the length scale can start from let us say one angstrom, nanometer, 10 nanometer, 100 nanometer, micrometer and then let us say millimeters and similarly on the time scale you can start with a picosecond time scale, nanosecond, microsecond, millisecond, second and so on. So this is kind of your modeling space. So when you look at a CVD system for making nano materials you can actually subdivide it into 4 regions. The region that is characterized by the longest length scales and the longest time scales is the continuum regime or the macroscopic regime. So this describes what is happening in the bulk of the reactor, how the reactants are getting fed into the reactor, how they are getting distributed in the reactor, how the products are flowing out of the reactor, how the velocity gradients are happening, how the temperature gradients are being set up and so on. And this is what again we describe using continuum equations, conservation laws, constitutive relationships and so on. Now the second scale here which essentially starts in the sub micron region and extends till tens or hundreds of a micron is called the mesoscopic region also known as the Monte Carlo regime for simulation. And then there is the next region which is called the molecular dynamics regime and then comes the regime that is unique to nano materials. It is basically the quantum mechanical regime which is characterized by for example the Schrodinger equation. So how do these things again dovetail into each other? When you start talking about nano materials again at its smallest level we are talking about the assembly of a few atoms to make a sensible structure. At that length scale and at that time scale we are really talking about the interaction of electrons with atoms. The atoms are actually responding to a field that is set up by the quantum electronics that are present in the system. Essentially the wave nature of matter comes into play. So you have to go back to your, I guess it was your first semester physics or second semester physics where you are studying quantum mechanics and quantum dynamics to remember how we used to describe the motion of electrons and very, very fine particles such as atoms in the earth's electromagnetic field. And that is really the level of detail that you need to get into at that level. What you really care about is the electron density at each location in the reactor and on the surface. What is the gas phase composition in terms of ions and electrons that are present in the system? How are they interacting with the atoms that are present in the gas phase as well as on the surface? How are they either aiding or inhibiting the motion of these atoms? Again both in the gas phase as well as on the surface. What role are they playing in either promoting the attachment of these atoms to make larger molecules or in preventing the migration and attachment of these atoms to keep them at atomic level? So the focus is really at that level. You know you are really trying to look at how atoms and molecules are interacting with each other and with the electrons and ions that are present in the system. So when you go from here, I mean these three length scales we essentially discussed in the earlier lecture. The molecular dynamics range, again what we assume there is that the kinetics of motion of the molecules is the rate limiting factor. So in this range, in the molecular dynamics range as the name suggests, you start studying the dynamics of individual molecules. You model a molecule as a very small particle. So we essentially apply the Newton's equation of motion. You look at the force on a particle, relate that to the acceleration of the particle for a given mass. The force itself is a result of various potential energies that are present in the system starting with intermolecular potentials and so for example the van der Waals potential plays a role. All the electrostatic forces, the contact potential forces between a particle and a surface, all these molecular as well as electrical forces play a role particularly in this regime. So you essentially characterize the acceleration of that atom or molecule or fine particle I on the basis of all the forces that it experiences and essentially solve the equations of motion for discrete particles. Again what you are doing is modeling molecules as particles and essentially assuming that the motion is deterministically decided by the Newtonian mechanics of the system. So for example if two atoms collide, you look at it as a hard body collision and you use conservation of momentum principles to see how they are going to move in the field and so on. So this is a slightly expanded view. You know here we are trying to look at atom by atom and here it is more of a, well I would call it molecule by molecule. The third regime is what we would say is an aggregate by aggregate model. So in this regime you have now formed films or aggregates that are of the order of a few microns and you start trying to understand how these aggregates behave. So typically this is done using Monte Carlo simulation techniques using probability theory as well because the laws that govern how attachment of individual atoms and molecules to form larger aggregates happens cannot be deterministic in nature and in the Monte Carlo simulation methods you acknowledge the fact that there is a probabilistic element to it and in fact the solution that you come up with for the position of the atom or molecule at any time or the size of the molecule at any time is not based upon deterministic principles but on stochastic principles. So in this regime you are essentially adopting stochastic techniques again with a focus on the substrate or the surface on which these phenomena are happening and finally you link all this. So the first three stages really describe what is happening on the surface or the substrate on which the deposition is happening and then you link it to the reactor through your continuum model or macroscopic model. So in terms of scaling this is how we want to look at it but in terms of actually let us say that you want to predict how a nano sized feature is growing on a substrate. How do you do that? There are really two approaches. The first approach is the hierarchical approach and the second approach is called the concurrent or hybrid approach. In the hierarchical approach what you do is you first solve the system at its highest level. You take the output from your simulation at the macroscopic level, feed it to your mesoscopic model and then again take the output from the mesoscopic model, feed it to the Monte Carlo simulation, take the results from that and feed it down to the nano scale model. That is called the hierarchical approach because we essentially follow a hierarchy of diminishing time scales and length scales. The other approach that can be used is to solve all of them simultaneously so that you essentially set up an iterative loop where you assume that it is not just a, I mean the hierarchical approach basically says it all goes this way that you can solve that system first then solve this, solve this and then solve this but in reality that is not necessarily true because what is happening at this level can influence what is happening at this level and so on. So the concurrent model essentially introduces a two way coupling and makes it into a iterative process where the final closed form solution for the system is obtained after you know setting up an iterative loop and obtaining a self consistent set of parameters for the system. In other words the growth of the nano material at this scale cannot violate what is going on at the continuum level. Now is it possible for such couplings to happen? I mean can this process influence that process? Well as an example you know you can take things like surface roughness. The roughness of the film that you are growing on the surface is very much dependent on what is happening at this level because when we talk about the surface roughness or surface asperities we are typically talking about scales that are of the order of sub nanometers particularly on highly polished semiconductor devices and so on. And the presence of even a few nano sized particles on a substrate can have a significant influence on the measurable roughness of the surface. It also has an effect on the 3D profile of the surface. So if you have a good AFM you can actually see the presence of the nano particles on the surface and you can see how for example flow around the surface is affected even by the presence of these very small nano sized particles. So certainly at the boundary layer level the presence of even atoms can have an effect and so that can feedback up the loop because once the boundary layer gets disturbed then that has an influence on what is happening at the exterior or the mainstream because of the coupling at that level and so this effect can certainly propagate the other way also. Now again the pros and cons obviously the hierarchical model is very simple simplistic you might say. It just basically splits it into 4 different problems and solves them sequentially whereas the concurrent model introduces a much higher level of complexity by insisting that all these equations must be solved simultaneously. One of the major applications for nanotechnology these days is carbon nanotubes. Nanotubes are being used for various purposes particularly as filler materials. They provide huge property enhancements for things like stiffness, elasticity, hardness of materials, conductivity of materials. So they are very attractive nano substances to try and make. Now when you look at these carbon nanotubes and you try to model how they are formed clearly you know the modeling has to start at the nanotube level because it has a certain very distinct shape to it and you have to really have a model that can predict the shape distribution of the nano materials that you deposit. So that is really getting down into the nano substance level and then you have to link it to you know how do you make it happen? You know what kind of carbon source do you provide? What type of deposition conditions do you provide? Is it even possible to make a carbon nanotube using CVD techniques or do you have to use essentially molecular self assembly type of methods? So to understand that you really need to be able to do this as well. Now people have developed models to simulate deposition of carbon nanotubes using CVD techniques and these models predominantly tend to be hierarchical in nature. They have really not even tried to deal with the complexity of using a concurrent model for that. Now on the other hand if you look at a silicon film formation, silicon deposition you know the first layers of silicon that form are actually of nano dimensions. They can eventually grow to micron dimensions and so on but there are applications increasingly in microelectronics where you know the volumes are shrinking while at the same time the functionality is increasing where you are requiring silicon particularly epitaxial silicon of the order of few nanometers. So when you look at the deposition of epitaxial silicon at the nano scale the modeling tends to be more concurrent in nature and the reason one reason for that is you know the chemistry of how silicon films are formed is quite well understood. It is predominantly an inorganic chemistry, the chemical reaction sequences have been well characterized, well understood and so on. So it is actually possible to try and develop a hybrid model or a concurrent model for linking the growth of nano dimensional silicon films on a surface to the reactor conditions. On the other hand when you are talking about making of carbon you are typically using a precursor that is organo in nature and the number of reactions and the reaction sequences are not as well characterized in that class of materials compared to the inorganic materials. So given the lack of understanding of the gas phase chemistry per se it is just too difficult to try and develop a concurrent model. So what people have tried to do is develop a simulation model that will predict the formation and growth of the carbon nanotubes and then empirically fit the data to the model and have enough tunable parameters in the model so that you can try and match your experimental data as closely as possible. Usually these are trainable models where you as the model will essentially self correct based on the match or mismatch to the actual experimental observation. So it is you know more like a neural network type of a model that people normally use. So it is not as fundamentally based as the models that describe CVD I mean silicon CVD but for practical purposes they do the job. So it is an interesting dichotomy that depending on the nature of the CVD film you can choose either this approach or this approach. I think to some extent it also depends on your computational capabilities and facilities because to do this you need a supercomputer you know high performance computing, cluster computing all of that is needed whereas this can probably do it on your PC these days for a good workstation. So the level of computational details involved is clearly much lower in the hierarchical approach. Now one of the challenges again in trying to make nano materials in a CVD reactor I mean if you look at this nano is a very small size but a typical CVD reactor tends to be quite large in terms of its dimension. So there is automatically a mismatch you know that you are trying to make very very small materials in a space that is quite large. The problem that introduces is that unless your CVD environment is very tightly controlled a perturbation somewhere quite remote from the surface on which you are making the nano film can have a significant impact on the quality of the CVD of the nano product that you are trying to make. So when you are trying to make a nano material the preferred way to do it is to have a reactor that just essentially fits around the substrate. Try to minimize the amount of free space. So techniques such as you know the designer molecule fabrication techniques such as the molecular self assembly enable you to do that. You can build a control system just around the substrate where you are trying to grow the film and have very precise control over how the film is being formed or how the nano particles are being laid out. But you really do not have that luxury when you are trying to make you know larger quantities of nano materials. So it is essentially a trade off you know when you are making nano materials can either have extremely well controlled quality or quantity. It is very difficult to design a reactor in which you can do both get to tons per day or more of production while still maintaining the precise control over the structure and morphology and the other characteristics of the nano material. This is where the top down techniques have a certain advantage particularly if you are not requiring highly complex molecules to be produced for your application. For example if you are just doing catalysis let us say you want to use silicon oxide as your catalyst. The main parameter that you are interested in is the surface area. So you can take you know micron sized SiO2 particles and fairly quickly reduce size to the nano dimensions. You would not be affecting the crystal structure too much if you do this carefully but at the same time you are not going to be adding any functionality either. But if your entire purpose is simply to increase a surface area per unit volume of catalyst material or per unit mass of catalyst material then top down methods work very well. The bottom up methods whether it is molecular self assembly or CVD are required when you have to build for example layer by layer functionality. If you want to make a nano dimensional film where at even at sub nano meter levels you require a certain gradation in the composition of the film or in the structure of the film then you cannot do that using top down techniques. So you have to use the bottom up techniques and that is where CVD in particular can come in very useful. I mean you can do the same thing with molecular self assembly but it is going to be much more difficult, much more laborious, time consuming, expensive and so on. Whereas with the CVD process we know how to do this. We know how to control the conditions of the reactor to be able to deposit films of known composition and thickness. So it is very easy to do a layer by layer build up of a nano dimensional film simply by exercising controls over the operating conditions of the reactor. And that is the reason why CVD as a technology is now increasingly finding use in the synthesis of nano materials. It is kind of a good compromise between the manipulation at atomic level and you know simple top down type of techniques. It is a methodology that gives you virtually the same level of control as the molecular assembly techniques provide while at the same time minimizing the cost and complexity of designing and operating such a reactor. So nano technology and its interaction with CVD is a subject that is really emerging now. There are a lot of papers that are being written on this subject and you should try to read up and catch up on some of the new things that are happening. I will forward some papers as well to the class so that you can take a look at them. I think that brings us to the end of CVD applications that I wanted to discuss as part of this course. So we will have one more lecture where I will try to as I said give you an overall picture of what we have covered in this course and get you set up for your final test and so on. So any question on what we have discussed today? Okay so I will see you Monday then.