 We were discussing CVD, we also said that there are number of ways in which films can be deposited using chemical vapor depositions, we have listed this number different things. This is last time we did and in VLSI we said we need to have metal films to be etched, the dielectric films to be etched and also alloyed to be etched, okay. Also we need atmospheric pressure CVD to deposit epitaxial films or growth, that is called grown films though it is actually depositions. The major important parameter which decides the growth or depositions is decided by the thermal budget. How much is, what is thermal budget? In a process whatever is the furnace or any other anneal cycles you go through, the time and temperature what you use is called thermal budget because it consumes wattage, okay. So is thermal budget also decides which process to use because in processes the last end of processing is actually should have lower thermal budget because earlier process should not change. So thermal budget is a very crucial factor in decision where to use which kind of CVD and of course quality of films to be etched after quality forms of the film desired to decide the choice of process. We discussed last time both APCVD and LPCVD, okay. So we have gone through all of it, I just want to show where we were and we were looking for the first process which is the atmospheric pressure CVD in which in a system, in an APCVD system you have a heater on which wafers are kept. This need not be on the conveyor belt, it is good if they are because then you can have larger throughput. The basically, may I have your attention now? You have a heater on which your wafers are kept. This temperature normally is around 400 degree centigrade plus, okay. Here you introduce first nitrogen for the clearing all the gas stream here. Sometimes you also pass through hydrogen to etch out few things, that is called reduction and then you actually add the process gas. In most cases it will be size, if it is silicon based films then it will be silane or trichlorosilane, dichlorosilane or something of that kind. What is the procedure? The procedure is reactants must reach the surface of the place where they want to be deposited and they are kept horizontal. This is important fact, they are kept horizontally. From ambient gas stream, the reactant must diffuse through a boundary layer which we will discuss later. So there will be a thin boundary layer on the wafer and reactant must, oh yesterday I showed you the film, sorry. There will be a boundary layer, stagnant layer as it is called and any new reactant has to diffuse through this to actually come and sit on this. So this is an important thing which decides the mechanism of growth or deposition. Then what happens next is this is what we did last time so I am now going, this is what we start now. Once they pass through the deposition this boundary layer they are now at the surface of the wafer or substrate and there these reactants must get adsorbed to stick there. Surface reactions then can occur and remember the surface reaction is decided by the heater temperature. At that heater temperature wafer are now held. So the reaction must take place at that surface temperature which is typically around 400 to 700, 800 degree in most cases but in epitaxial as I say it may be as high as 1100. Deposition occurs conformally. This is most important wherever the step is there it will climb because gas can flow everywhere. So it will conformally actually deposit, it will also have some something coming out of the wafer which is called emission and some may redeposit itself. So there is a combination of deposition, redeposition, etching or what we call emissions and at then some steady state is achieved. So unreacted reactants and byproducts are dissolved and they are actually removed out of the system by vacuum, vacuum pumps. We can have a kinetic subsets deposition similar to Gro-Diel model and we last time discussed that this is what the model is something like, this is what we did I am just trying to go through quickly what we did and we say there is a gas stream flux and there is a reacting flux at the silicon and in one is proportional to the Cg-Cs through a mass transfer coefficient and in steady state the reacting which is happening to proportional to Cs must be equal and then we derived this formula and we say okay Vy which is the growth rate which is F by n, number of atoms incorporated on surface is n, then Vy is ks, 1 plus ks by hg to the power minus 1 Cg by n and then we also discussed last time that the gas reactant in a total gas system, whatever in a gas system and how much is available at the reactant surface, the ratio of that is called the mole ratio and that we defined as Cg by Ct, Ct is the total concentration in the system. Based on this we finally arrived at an expression which is ks hg upon ks plus hg Ct by n into Y, Y is the mole fraction. We also looked into last time 2 cases, we say one possibility is ks is much smaller than hg, the other possibility the hg will be much smaller than ks. Of course there will be a point where ks will be almost equal to hg and that is called the critical shift point. At that temperature one process shifts over the other. So something below higher temperature it will be mass transfer and below that temperature it will be reaction rate limited. If you see that graph which I showed you earlier figure, so you can see from here any higher temperature it is constant with temperature means mass transfer limited and below it is function of temperature which means it is reaction rate limited. Okay and we then discussed this is all that we did last time I am just hurrying through them. Now we say the reaction rate limited constant ks as we know it is always function of activation energy and temperature, k0 is called its pre-exponent constant and please remember ks is a very strong function of T as one can see from here and the activation energy for the process. However we also said that the mass transfer coefficient hg is directly related to gas stream pressures and its diffusivity through the boundary layer. Since hg is a very weak function of temperature though it is a function of temperature but very weak function of temperature it is normally treated constant. Please remember hg is a function of temperature but it is very weak function of temperature so hg at after certain temperature we say it is almost constant. Though this can be verified by numbers and see how much it is but normally this assumption is fair enough. What we are trying to say essentially if the gas is enough there how much is reaction will be decided by the reaction rate but how much is made I can react as much but you were not providing me so it is mass which will allow me to limited. So at higher temperature what happens that the reaction rate is very high so anything can come and react with it but the amount gas will now decided by how much you push okay. So is that physics clear that at higher temperature reaction rate is very high so anything any amount you push it may react but how much you push will be limited by that is the hg term okay. Whereas at lower temperature you may have any amount of gas received at the surface but how much it can react will be decided by the temperature which you have at that point. So this shifter that is very crucial at some temperatures the reaction rate limited process goes to mass transfer limit how much gas is available for reaction okay. So one can clearly say at lower temperatures ks is much smaller than hg and at higher temperature obviously therefore hg is much smaller than ks okay. Now if I look at this term I think you wrote down if you look at this with ks is much larger than hg and if you go back into that expression you can see that ks term will vanish 1 upon ks is very large as small so it is only hg times something will appear there so it is only mass transfer coefficient hg which will decide the growth or depositions whereas if ks would have been much smaller hg 1 upon hg would have been higher as smaller and so that term would have gone and you will be only proportional to ks times the other parameters. So ks is decided by temperature hg is decided by available gas stream pressures or available molecules there to react okay. This is exactly what we differ between apcvd and lpcvd this is what we are trying we want to make a certain process only temperature dependent so what is the condition I am looking for that at lower it process is normally at look relatively lower temperatures where hg is much higher than ks is that correct hg is much higher than ks and then we say it is only temperature dependent. So to make hg higher we will see now that the pressure as lower the pressure you do hg will start increasing so at lower pressures hg could be made higher than ks much higher than ks so in low pressure cases it will be decided by what temperatures once I say it is only temperature dependent what kind of wafers wafers I can hold I can hold vertical wafers why because I only look for their temperatures in a furnace I have hundreds of wafers but they all can be held at same temperatures 200 wafers or may be even 400 depends on the size but all of them should have same temperature as long as the furnace temperature is constant the deposition rate will be limited only by the temperature constant there if delta t varies just growth rate will also or deposition rate will change otherwise any number of wafers can be deposited in one go so it has a large through boots normally polysilicon is deposited at low pressure so it is called LPCVD whereas most films like SiO2 or Si and silicon nitride which are mostly amorphous films we are not really looking for structure there is no real requirement for good structure is amorphous anyway so I can deposit at much lower temperatures but then what I should do available gas should be sufficient for the thickness to grow okay and therefore all mass transfer limited way condition wafers should be flat because gas stream must come there and start depositing in the case of LPCVD wafers can be held vertical so what is the first difference between pressure changing that the throughput of a all mass transfer systems or low temperature system is very low compared to of course the belt system which I showed you say yeah I do increase it I push it as fast as I can but after all I have to grow film at must some retention time I will have to give for the gas so when it finishes that then only I will move out okay and the way it is done is that if there is a conveyor belt and as I said 12 wafers are sitting and gas stream is moving like this and we maintain some velocity we will discuss we may not discuss this so we know when the gas is flowing it is going all over okay so growth is varying every point in fact okay so the rate is so adjusted at this belt that at the end of the day everyone gets the same thickness so there is a belt speed which is very crucial to decide uniformities okay now among the most famous low temperature not low temperature the mass transfer limited cases is the growth of epitaxial films the word EP stands for as it is taxi means whatever texture so if you have a silicon wafer of some kind of 100 orientation or 111 orientation and I want to deposit silicon over it okay why do I want to do it because this EP layer I can deposit uniformly that is I can dope the wafer during growth itself so I have a p type substrate and I can actually deposit n type silicon itself of my choice of doping okay of my thickness requirement in a CMOS if you have seen there is a p well and then another I created another layer for a VT edges but this p well is sitting on a substrate okay so I can make instead of wells I can actually make two different epitaxial growths with thinness this is done in what we call the process called SOI process silicon in on insulators okay so silicon can be deposited also on insulators as well as silicon and these processes are very dominant now for the case of low power circuits SOI in particular the reason why SOI has not caught the what type applications and so this epitaxial growth is a very crucial requirement nowadays but earlier it was done only for the case of bipolar transistors we start with the collector n type okay and then we deposit or rather grow epitaxial layer of p type which is constant base why we wanted that because we any earlier we never used to control the profiles there were no implants so we want that doping of base to be constant so base is to be always epi growths okay and then I will diffuse the emitter in that and I will have an NPN transistor which is vertical okay which is vertical those days ICs were not used often it was only single transistor making and then we externally used to connect so in those days EP was major requirement for bipolar processes even now bipolar process do use EP but even in mass for ultra thin body of infrate also there is a thin layer we want to have it so which is also epitaxially grown okay. So there are processes for low power low threshold processes fast circuits low technology node we still need EP layers okay only problem with EP is it takes large temperature now this fact is very interesting because if you say that the process is related to temperature higher temperature because crystalline it is to be attained one way of increasing the temperature without increasing this environmental temperature is to do plasma okay plasma KT could be as 10000 degree centigrade can get to okay so therefore now EP layers may be probably grown by plasma processes rather than the normal thermal processes which is not I mean still some places when first process is itself is EP you do not have to worry the plasma is costly and difficult to control this in number of efforts can be deposited in one go so EP is cheaper in compared to plasma the throughput of plasma systems is always smaller than throughput for EP reactors so please remember also there is one figure which book may also given you since the gas stream is moving from say left to right in this figure since the what is our condition when the gas goes over the surface it starts climbing why start climbing because the pressure there forces the gas to go up so when you start here the gas outside may be what horizontal it is but when it touches the wafer actually it starts climbing but what is the condition I want to hold is this I want mass transfer coefficient which is given by diffusivity of gas to the wafer through the boundary layer by the thickness of boundary layer so what is when I can make HG constant when delta is constant is that clear DG is diffusivity which is known for a gas 2 gases but what how can I make HG constant if I make delta constant what is delta the boundary layer sitting if this moves away that means delta will start increasing delta will start increasing but in my case then HG will vary so what I do essentially is I put the wafer on an inclined plane since it climbs here so the actual difference between wafer and the gas stream is roughly delta all around this is all EP reactors you find as susceptor which is kept at certain degree which is around 28 degree climb okay so this essentially makes this delta constant and if delta is made constant HG can be made constant and if HG is made constant the growth is proportional to HG is that clear since it is a mass transfer limited process HG is constant and therefore one can see higher temperatures so no case related system so everything is mass transfer and therefore EP grades are normally done at higher temperature to get crystallinity and also we must maintain the angle so that uniform delta is achieved therefore uniform thickness is also achieved is that clear this is something which is very important in EP layers if you want to dope this EP layer okay so there are certain things if you do you are write down I will just give you what is how can you dope a film that is I was passing silo in for silicon so I should also pass di borane for boron doping I mean pass phosphine or arsine during this so that mixture of silo in and arsine will make N N plus kind of this the gas mixture decide the concentrations is that clear how much ratio you keep will decide N or N plus P or P plus okay that is how it is adjusted so all EP layers are uniformly doped this fact has to be always understood what is the difference between normal implant or anything no profile it is uniform doping okay and that is the feature of an EP growths okay and I really said where and where else the EP are actually used now for single crystal silicon solar cells one of the method of making films of EP layers on substrates which is not necessarily silicon but silicon dioxide is essentially EP growths because they are relatively cheaper compared to crystal growths okay I am not saying they are cheaper related to and they can we have a belt system in which films of silicon can be deposited and then you can make solar cells which are relatively cheaper is that what layer so do not think cheaper relatively cheaper so solar cell people normally use all belt system because they want thinner films they only want one junction and they want larger throughputs cheaper throughputs so most of the time films are used in their techniques in VLSI we need some time depletion regions to expand which means I need region there where it can expand so I will actually expand N deeper P sideways so that my depletion layer goes down okay so there are difference in device characteristics and mass devices and in the case of solar cell which is normally a normal PN junctions okay typically if you are written down this most cases the process which is does not use silane then it normally use either use silane or normally silicon tetrachloride which is actually liquid is used in most often the reason is whereas silicon tetrachloride is very less toxic so 2 achievements are getting the liquid which I use can be vaporized by 100 degree centigrade so I get a silicon chloride vapors and also it is less toxic whereas if I use silane it is not that toxic or very little toxic but it is extremely highly flammable material little bit of heat extra and pressure increase it will blast okay. So we avoid it and it is not that we do not use it we use silane as often as possible but just for the heck of it we say why tetrachlorides so silicon tetrachloride react with hydrogen to form silicon and HCL HCL is taken out one can see what many reactions actually takes place there is R-scene actually decomposes into R-scene and 3S2 R-scene gets attached with silicon at this temperature there is sufficient diffusivity silicon plus SiCl4 can form dichlorosilane dichlorosilane will further react with hydrogen to create silicon and HCL and this of course there is another process which is trichlorosilane based process. The cost of process is decided by what dichlorosilane trichlorosilane I use the for example to grow a polyfilms the if you are using a dichlorosilane it is the costliest process if you use trichlorosilane it is cheaper if you use silicon tetrachloride is even cheaper but since the throughput will be less with silicon tetrachloride so normally trichlorosilane process are used in crystals or polyfilms okay all ingots you have seen in crystal growth for actually using trichlorosilane process okay. So you add arsenic you add phosphine you get phosphorus you add diborane B2S6 you get boron so depending on the mixture ratio one can decide the concentration and thickness of the film thickness is decided with the time for which gas stream moves in okay so before we quit this area I may just tell you some maths I did which may be interesting for you okay I said very casual time okay if you reduce the pressure HG increases so I will now prove that HG does increase this in LPCVD the pressure is reduced typically it is around 300 millitars of pressure 300 to 600 millitars of pressure which is maintained some LPCVDs are also been done at one milliter but there are different processes some other day okay. So obviously we know the diffusivity of gas is proportion to the total pressure of the gas stream there okay so we now want to look for this pressure dependence and as I keep already saying I want to have HG increasing such that I became I move away from this and I get some kind of constant HGs okay this is what my game is if I change the pressure HG increases and then if you plot that same characteristics HG becomes higher and higher and finally becomes flatter at lower temperatures okay that is the idea. Total temperature of poly growth is around 300 to 700 degree centigrade 300 is not good films why 300 is not good film polycrystalline requires some crystallinity so lower the temperature it will more amorphous material so you will annihilate once you annihilate you already increase the temperature okay so that whole purpose of reducing temperature may not be worthwhile okay. What? Anil can be done anytime anywhere in the furnace where we are doing you stop all gas pass nitrogen anil okay you can take it out pass through a RTP system or RTA system you can do anil afterwards anil is not a process which in general in situ is preferred because wafer takes out means you need to clean it again before you anil inside furnace you hold it actually okay so you change the temperatures or the way it is done is there is a different zone where anils are performed in the same furnace so wafers are pushed there by track the temperature is different whatever you set and pass nitrogen heavily on that anil whatever time you want okay and then pull out the wafers. Wafers I can stack 100 200 wafers on a rack is that clear for the HG limited what is the condition it should go on the surface so I have to keep wafers flat so on a susceptor how many 8 to 12 inch wafers I can keep whereas vertically I can have a millimeter gap between them and I can have in some 1 inch 1 feet around 100 wafers if you have 2 feet zone I may have 300 wafers okay so I get in the same gas everything so many wafers can be deposited in one go okay whereas in other processes at mass pressure they need to have mass transfers flat process numbers are smaller LPCVD okay there are 2 which I did not say but maybe you have asked it so the problem with all horizontal reactors is unless the temperature outside the wafer susceptor is very low there is a contamination of particles from the walls so which actually sits on the wafer so there is a price system every few this you actually stop the gas and push the particles so it is a difficult process to maintain actually so preference is always given to vertical where even if it falls hopefully it will not touch the all wafers every surface okay may fall down okay so normally preference is always given to vertical holds but wherever necessary there is nothing else I can do so I will try some best to see it is less contaminant but flat okay you haven't said so this is something more interesting I say I want to look for pressure dependence and I want to prove hg is increasing that is what I said I said dg is proportional to total pressure and typically it is found for the most of the systems which gas systems we use either dg is T to the power 1.75 and in some cases it is T to the power 0.5 so in this range roughly dg changes with temperature we assume and dg proportional to 1 sometimes and say very little change it occurs but it is strongly function of total the as I said dg is not independent of temperature that is what I keep saying so the dependency can be taken care through this term as well viscosity mu why viscosity has appeared suddenly anyone what else has meant as I have viscosity whenever the gas stream of fluid flows what is the mechanism of fluid flow it follows either the disturbed flow or it is called laminar flow laminar flow is decided by Reynolds number Reynolds number is OUL by mu okay so viscous means it will drag itself so it will move with lower velocity so stream velocity drops in fact okay so more chance of contaminations okay and more chance of boundary layer not properly adjusted so normally the laminar flow is adjusted through the stream velocity and as well as the viscosity okay this year I am leaving many much of the chemistry but viscosity is view is proportional T to the bar 2 by 3 and normally it is not a function of total pressure okay I will leave it to you to figure out PV is equal to NRT is the formula ideal gas law says maybe I can write here for you PV is equal to where R is the universal gas constant N is the substance or amount of material reactant you have per mole okay and T is temperature in absolute temperature V is the volume of the gas P is the pressure now think of it this can be converted to P is equal to actually you should write R some dash you should write this is modified universal gas constant R dash which is essentially R by M okay some other day some gas law product or so whatever is PV is equal to NRT formula which who actually gave this formula Clapeyron actually created this equation after learning that pressure is proportional temperature and volume is also proportional he actually made this equation together from Boyle's law and the other law okay so please do not think it is actually Clapeyron's equation which is PV is equal to NRT okay but somehow it always got to Boyle and everyone gave credit everything to him okay but just to be history I mean tell you okay so rho is atomic mass density rho is proportional to PT and rho is also universal proportional to temperature gas in velocity U the net velocity of gases going through in the system is not a function of temperature it is only how much pressure backside I push the gas in okay however it is function of pressure to some extent T to the power minus 0.9 many a times one can take its universal proportional to PT now let us do the game I said okay it is the pressure so one atmosphere pressure that was go thousand time reduced okay so 760 Torr to 0.76 Torr thousand times that is 760 milli Torr 8 Torr say so one atmosphere pressure say atmosphere see by comparing with that because earlier process was atmospheric pressure CVD so 760 Torr thousand time pressure it is for different gases diffusivity is different so it varies somewhere for each so in normally we may not even consider that but I just show you DG is not independent of temperature earlier made a statement okay DG is a function of temperature so okay in that case assuming that temperature dependence is smaller DG 1 by DG 0 is the ratio of pressures if I assume it is PT DG is proportional to only PT first term so DG 1 by DG 0 G 0 means atmospheric pressure this is at the thousand time less pressure so DG 1 by DG 0 is thousand times that is what you reduce the pressure so ratio is this one upon and a so U1 by U0 is some proportionality I say C P1 to the power 0.9 C P1 P0 to the power 0.9 so it is P1 by P0 to the power roughly it is 500 so the ratio is 500 for U1 by U0 DG 1 by DG of course this I am neglecting temperature we can add that if you wish then similarly row I know is proportional to this so I say row 1 by row 0 in this assumption is temperature is held constant only pressure is reduced so I am only right now looking pressure dependence so row 1 by row 0 is 10 to is that point clear temperature is held but pressure is reduced from 1 to 1 atmosphere to 0.76 Tors so row 1 by row 0 which is proportional to PT so it is now thousand times okay so now I what this material science pseudo chemical pseudo whatever it is person should know is the Rinal number which is described as Rho UL by mu L is the length over which gas passes through susceptible lengths okay so Rho is the atomic mass density U is the steam velocity L is the length which is constant for a process and mu is the viscosity so if I take the ratio and I want what kind of flow why I want laminar because if the particles themselves collide then there is a diffusivity will be various and different angles I want everyone to go through down so I want laminar flow and that is decided by the Rinal number typically less than 8000 is expected RL 0 by RL 1 is Rho UL by mu upon Rho UL mu is independent of pressure we said so it is Rho 0 by Rho 1 U0 by U1 L is same for both susceptible lengths so roughly it is 2 if we require figure it out the boundary layer is proportional through this to Rinal number is two third L by RL root RL this boundary layer is related to laminar flow and susceptible lengths okay so it is two third L upon root RL okay so now delta 0 is two third one of L upon RL 0 delta 1 is two third L upon under root of RL 1 so delta 1 by delta 0 is RL 0 by RL 1 under root which is root 2 because we got this number 0 2 so delta 1 by delta 0 is root 2 now Hg 1 by Hg 0 is dg 1 by dg 0 delta 0 by delta 1 this is 1000 this is 1 upon root 2 means 1.41 so Hg 1 is 700 times Hg 0 of course these are not exact numbers but the proportionality constants are not true in all pressures all temperatures but assume right now the idea is to show you that if I reduce the pressure 1000 times the mass transfer coefficient will be as large as 700 times the atmospheric pressure so it just by in a furnace if I reduce the pressure I can increase Hg and if I can increase Hg then what is the process will be limited by temperature so vertical stack and everything is possible uniformity is temperature dependent how good temperature you maintain okay in a vertical you cannot create laminars is that point clear the gas has to go like this like this each wafer laminar is always attained when it you are in a surface in a vertical the gas cannot be streamlined very much because the wafer size and the tube size does not allow laminar systems so kaisa gas jata hai ek wafer ki andar jayega phir baar aayega phir jayega so it is unlikely to create any laminar flows okay oh no it is not that is what I say because it depends on the size of the tube and size of the wafers laminar part lengths susceptor lengths so Reynolds number changes with the size okay yeah yeah so you have to worry all the time how to maintain laminarity it is not very easy but they are not dependent on mass transfer so you need not worry I am only depending on the available gas is not my choice I want temperature any amount of gas will react temperature will decide how much to react is that clear that is the game I want to see it is only temperature dependent so I must increase my hg somehow so I figured out if I reduce pressure I will increase hg okay so this why I said you earlier that how from when I said hg will increase if I reduce pressure I just did some calculations there are some small machine which you cannot find but truth is still seen I can also do similar thing for temperatures and I did that if you are written down please so is that proof clear that why if I reduce the chamber pressure the mass transfer coefficient will increase as much if not 1000 but at least 700 800 times which means case dependent depositions will start because hg will be much larger than case so that is the purpose of doing all this okay so you can reduce the temperature you can increase low you reduce the pressure and then you are always temperature limited processes okay this is a method which is in mind so anywhere I can use this oh I must reduce this I will get this they have to even oxidation but they are also it is thermal limited they are also we the process when I start papers I said case is much smaller than hg so anytime hg is larger I will always be safe luckily in the case of this this silicon dioxide allows you in the oxide grows the diffusivity is another term which is helping you okay so there is a that term is not available here okay so if I do it it is imperative that hg increases when I reduce the pressure similar analysis was done by me and now I do not want to show every term I only give you the data dg is proportional 1 upon p dg is proportional t into 1.75 the viscosity is t to the power 5 actually it varies between t to the power 0.66 to 2 third many people have different versions raise proportion to p and raise also 1 upon t and if I now temperature is lower down by t and if I do the same analysis which I did for pressure dg every calculation I perform typically I get hg1 by hg0 root 2 times so exist so even if you reduce the temperature by half your hg will be at least double is that one or at least one and a half times it will be if you reduce pressure it increases enormously but with temperature also it actually increases but you can see I told it is much smaller dependence so it is it does change with temperature but not very strong functions is that point clear when I made a statement I just wanted to show you the changes much smaller even if I have the temperature from 900 if I go to 450 hg is hardly increased by 1.4 one and a half times whereas in the pressure if I reduce 1000 times I am just going away 1000 times impression this hg so this fact is actually used in all LPCVD APCVD every system which I actually use in my this is in my mind if I do this this will happen is that okay and I have done all this analysis for this also I am now giving you final result you do same thing which I showed you for all terms make division rule you everything and figure it out whether it occurs which occurs and this is my result so you can also do it. No there is no delta no hg is so high that means delta is very small that exactly is the point so almost all gas is available on the surface so there is no diffusion required because gas is right now so much available it is only how much it can react decides is that clear hg higher essentially means there is no boundary layer very much the gas velocity is good enough to actually make system only temperature dependent that is the way we do it so typical pressure I said you know LPCVD normally for poly silicon is done around 300 to 600 milli torsors pressures 0.3 to 0.6 tors is all that pressure we maintain temperature is typically from 400 to 900 sometime but 750 is all that we try same way you can do silicon nitride instead of poly if you want you add ammonia along with silicon okay and you can do same LPCVD okay so you have depositions of silicon nitride any film can be deposited if as long as mixture is made at that pressures of course these are figures which are given taken from circuittoday.com website I do not want to actually just I want to show you the actual there are number of kinds of there is a horizontal reactor there is a vertical reactor there is a cylindrical reactor for repeats the advantage of each system is the throughputs are changed like there is some in this actually there is a hexagonal system this cylindrical system has hexagon or octagon so each may have 6 wafers this and it rotates okay so there are methods are improving throughputs this is available as I say please note down this side which is www.circuittoday.com so this is a standard EP reactor also okay I forgot but maybe most EP reactors do not use thermal heaters okay what is the purpose of this I just want to make it clear which I did not so far I only want temperature of the susceptor to increase I do not want the whole tube being has higher temperature so if I make a resistive heating what will happen the whole furnace tube will get heated is that clear so I now want only the place where I want reaction to be heated to a temperature of my choice and the rest should be cool it is called cold wall reactors then how in a heat this wafers the only possibility is through induction heating so this is a graphite susceptor inside which there are rods okay which acts like a and the resistor acts graphite acts like a resistor across a transformer coil equivalently okay and the outside there is a transformer other coil you actually create RF inputs there at high frequencies this induction heating will allow this susceptor to heat okay so what will be heated only the susceptor of course it will some temperature will increase of the because it will be as some way but most of the time only susceptor is heated but the walls are cold so these are called cold wall reactor cold wall walls are at room temperatures same thing is possible in plasmas the walls are cold but the place where you want can have higher temperature so this whole procedure is associated to make it cold wall systems is that clear okay there are other same from the same site I had taken just figures of others this is LPCVD this is APCVD you can see vertically stack wafers this is all AP LPCVDs have vertical stack wafers this is of course belt system for atmospheric pressure then there is another process which is called PECVD plasma and ANCVD now here the susceptor cap is here which is heated through a half sample holders these are wafers okay this is a gas inlet and this is RF input at the other electrode using a plasma I can actually create deposition of any gas which is analyzed is that clear so this is a gas enhanced CVD this is called low temperature PCVD which has actually a normal zone and then you actually create at the some point internally plasmas on the wafers okay so it is a localized plasma reactors this is very costly system which is actually called hot wall because the others are actually in furnace so this whole track is pushed inside in a furnace so if this process allows you to make even crystalline silicon okay because higher temperature you get it out plasma actually allows you to deposit and it anneals because of external temperature so you get actually EP growths okay so this reactor which is low temperature PCVD is used for EP growths why this EP board I said the higher temperature externally will anneal it out internally at low temperature it will deposit it out okay so it is one go or in same system you can create and since they are flats they can always be uniformly crystalline growths okay so this is one possibility just for the heck of it I will quickly read through EP CVD is mass flow deposition system low temperature deposition are not possible hot wall plasma could be possible that is what I said the biggest advantage of this is uniform growths or deposition larger die wavers and you need lot of gas actually you need huge amount of gas actually why you need larger gas because whole tube has to be filled up please remember many I do not know any anyone of you is doing any process in the lab and any time you are doing poly or anything please see that what is the tube size and what is the volume of that tube what is the volume pyre square into length at any time amount of gas should be sufficient that it fills up this volume otherwise they will be pockets and actually before my break because of stress okay so even if you are doing in a any others gas stream the gas requirement is very high depending on the size of the furnace 80 liters a minute to 100,000 liters a minute flow may be required so very huge gas flow and how much gas you will use very little few CCs 100 CCs okay and the rest will be just going away okay but that is the cost okay so APCVDs are costly in the gas systems waste the too much of a gas requirement is smaller but it has to be maintained because it has to fill it up all the time this is given in the circuit today you can read this this for example for LPCVD if you do APCVD there is a poor step coverage if you do LPCVD you have excellent coverage I already discussed LPCVD typical temperatures okay before we finish this area maybe I have one interesting figure for you okay there is something which I will put it on the web today myself this you should know but I know you do not know so I actually written down all units which are required in CVDs okay or anywhere else but at least more in CVDs or PVDs 1 electron volt is 1.6 10 to power minus joules which is 1.6 10 to power minus 12 hertz or is equal to 3.87 10 to power minus 38 calories or 23.5 kilo calories per mole 1 joule is 10 to power 7 hertz 1 gram calories 1 gram of calories 4.18 joules 1 watt second equal to 1 joule is 0.24 calories for pressure 1 atmospheric is 760 millimeters of mercury 760 tor which is equal to 1.013 10 to power 6 dines per centimeter square as a pressure or 133.3 Pascal's 1 dine per centimeter is 7.5 10 to power minus 4 millimeters of mercury 1 Pascal is 1 Newton by meter square which is 0.0075 tor 1 bar is 7.50 10 to power 2 tor and 1 bar is 10 to power 6 dines per centimeter square. So the units in each processes is different yesterday I did give you some kilo calories conversions okay because specification from thorough dynamics may come in one unit but for the operation and what units we are using we will have to convert to EV per mole okay. So whichever expressions I give I do define which unit I am having so and maybe I will get this we can get finally in exam also so that you must convert properly to the units is that clear otherwise that numbers will be awry okay. So this unit calculations bar to Pascal to tor is very crucial because some people may say 2.3 bars you will say what is this bar now okay so I just tell you that there are number of ways in which pressures temperatures energies are mentioned and therefore some conversions are actually needed okay. So this is taken from variety of my old notes I finally copied once for you okay so that you know now what is so if I see just some expressions say all these yes 6 months I am teaching 4 months different times I have different units in that expression because like current I want something in centimeter something something so I always but this actual data may come from other areas okay so they may give something so much dimes per centimeter now you are asking force you want now in terms of pressure you must convert okay you must convert okay is that point clear okay. So this is a LPC video typical pressure is 30 to 250 Pascal's 300 can I roughly say 1.3 so it is okay 10.300 to something like Pascal's tors millitors 5% accuracy is only possible advantage excellent uniformity very large number of wafers can be kept large throughputs large loads larger dia wafers can be kept because any size as long as tube can take that size is fair enough for a 8 inch wafer what should be the tube dia then it will never fit do you know why because the rack this is a cylinder circular tube the rack will take one quarter of that is that clear then you have an 8 inch then you need 2 inch clearance so on end average foot 12 inches minimum required for 12 inch wafers okay so you can see 8 inch ke liye 12 inch ka tube 12 inch ka tube ke aine pi r square bada aine aine length aapke to bade hai so gas flow bada dia so if I work on 1 inch wafer I can get same result but if I had to transfer to this I need huge system lot of cost okay this is the problem with our labs we cannot shift to higher certain they say wafers of 3 inch are not available so now convert it to 6 inch they may say so but here all tooling if I had to change next 100 karar ke der the lana mujhe bine pata okay anyway the deposition has only problem is low deposition rates generally and they are toxic gases are involved wafer as a quarts ka rack reta actually us ke upar slot slot me dalte slots reta us the wafer us ke under adat pi how the wafer so there is a slot in see in the rack a flat rack reta hai jiske niche kar part reta hai jo hoi karta hai us meh aise slots cutter reta wafer hare wafer ek ek slot meh jaata hai you cannot connect to each other na gas ke si jaayi so gap te rakh nahi padega na do slot ke beech meh wafer jaane ke liye gap te rakh nahi padega this is another system which we discussed plasma and an CVD normally it has a rotating wafer holder then you have a processing glass there is a RF system we will discuss this little later in etching so I do not want to write now discuss advantages of PCVDs are that pressure vacuum should be as low as possible as higher at lower temperature it produces uniform layer multiple use of this system it need not be one way any kind of wafer can be used and it does not spoil very much because it is only plasma ions strike only targets. There are many more variables in the plasma system RF ka matching ho na chahiye distance ke negative glow kitne door rakh nahi there are many requirements to fit in and it is very costly. Sab sab bura hai ki uska throughput is come. These figures do not draw because these are available on circuits no ya it is all on the circuit today this this model which I am showing is for do-corning. Ya aisa sab system joh hai ab ko dikhate ke actually dikta kaisi? This is your PCVD. This is from plasma rods which is system 1200. So tikh hai asab system 100. Ya exhaust pipe hai you have to connect it out through a scrubber toxic gap baar nahi jaana chahiye bhot ko chiche karna patiye but this is the system which is possible. The last but not the least of these process is very important nowadays and which is called atomic layer depositions. The last but not the least important is atomic layer depositions very very important process these days because of all lower node technologies. This is essentially you can see there are two gas stream entry which is called one and two. One is actually used for some purpose to I will just discuss that. Then there are wafers okay. Then there is a susceptor, there is a some sort of system in which gas is passed and deposited. Now there is something which we do actually. First the reactant gas is entered. This is called first process, step one. Pulse of reactant gas which is now at the surface of this substrate. I allow gas to come in okay. So it react and it is temperature everything is adjusted. So it actually flows over it okay. So some atoms are dissolved there. We will come back to it. Pulse excess unreacted gas. Then what do I do it? I stop reactant but pass purging gas. Nitrogen in many cases or gone in some cases. You can see here the atoms are here and the rest of the gas is taken out. So it is called purge. Some layer is dissolved and the other part of the gas stream is taken out okay. So it is called purge. Then you have second okay. This reactant gas is also in actual ALD system is called precursor. So I thought I should write the name. It is called precursor. The reactant gas second time you have a precursor. This is called phase 3. So you purge gas again reactant gas. Let further reaction take place and purge it again okay. This you may have to more than once or more than 2 times for the thickness to be grown but this is essentially called purge. This pulse 1, pulse 2, pulse 3, pulse 4 system in which atoms actually can sit. First adsorbed, then react, then rest is going out and you get films okay. This is called atomic layer depositions, ALDs. Now if you want to add something to this film in the second pulse cycle reactant gas you can add another reactant gas. Like I want phosphorous to be added. I can pass phosphine through the second pulse I mean second pulse or fourth pulse okay. So I can add also anything along with that okay. But it is one cycle only one gas will pass okay. Why it is called atomic layer? Because atom by atom it sits okay and that is why it is called atomic layer depositions. Very important process of today okay. You can see here first pulse this comes in some of them sit then it is purged so rest going away. If they are more susceptible it keeps going. Second again you enter new ones then the purge okay. So every time you pulse you add reactant gas when you want to remove the after desorption you want to remove the rest so you purge it out. This layer by layer depositions which you can see here atomic layer depositions. So very thin films can be deposited using ALDs. The growth rates can be as low as 10 Armstrong per second okay. So it can be a mono layer can be deposited using ALDs. Accurate thickness control on a very wide range of materials large area uniformity operates at very low temperatures. No gas phase reaction is only adsorbed and then reacted okay deposited. Disadvantages of course is low deposition rate but that is what we need. In fact we need low deposition rates can be an advantage. I want a very thin films so I will use this. Reaction chemistry is quite complex. Poor crystallinity due to lower temperature. So silicon is not deposited for crystallinity. Films which are non-crystalline are easy to deposit in this matter. Organic materials all kinds of non-silicon devices can be made using ALDs okay. Silicon please remember this technology course was essentially highlighting silicon but that is not end of the world okay. Currently there may be 30% effort in non-silicon devices okay. Whether they will compete silicon in their next 20 years or 50 years is only God knows. People who are working there they are not interested whether it will be technology or not. They have publication so they will get it. But whether this will finally replace silicon is very much unknown and I am not very sure. Possibly yes okay. Is that okay? So I have given you all kinds of CVDs okay. I have also shown you PVD as well as CVDs. So any deposition can be made okay. So if I want to deposit tungsten what should I use? Tungsten fluoride is a gas which is available WF6 and you can go through plasma, you can go through sputter, you can go through anything ionic this and bombard that. So any material can be deposited by either PVD or by CVD. PVD of course not all but CVD almost everything okay. So CVDs are much more popular but why than PVDs? Because PVDs deposit metals very uniformly. So most metallization systems use PVDs. If you have a target of that material is best way to do it. This is our last sheet which I have discussed with you. Okay so we start with quickly some other area which is of more interest than the one which we did but we have now very little time to do it. Nothing can be done in VLSR unless you do etching. Unless you do remove some area then do some processing. So removal is more important many times in real life. The whole technology is decided by how old etching you do okay. That is the major worry though that is coming last as if it is released important but it is the important process in all of it. So there are 2 possibilities in which one is called weight etching which we have been doing many years and of last 15, 20 years we are working on what we called as dry etching. In dry etching also there are 2 kinds. One is plasma etching which is essentially chemical etching but no ions there dry. Then there is a reactive ion etching. The ions are present that can also chemically react okay. Of course third one is another dry one which is spotting spotter etch maybe I should add which is actually bombarding this okay. So plasma also has 3 kinds of etching possibilities. These both of them of course this is less. This is isotropic we will discuss this. This is very strongly anisotropic and this is also anisotropic but very low etch rates and much damage. Weight etching of course is isotropic. Isotropic word will come soon that is my job today to show before we leave. Typically as I said we need to etch SIA to silicon nitride, silicon, polysilicon, aluminum, titanium, molybdenum, tungsten, vanadium, copper. Then we need alloys and compounds like titanium oxide, titanium nitride, silicides of molybdenum, platinum, titanium many many materials we need to etch. And of course one which I did not write which we use etch every now and then okay. What is it? Potor resist or resist that is removal is every now and then. Every mask you go you have to etch that okay. Weight etching means solution based. Whatever you dip into solution and reaction will take place and etching will go. For example for silicon dioxide hydrofluoric acid is a very strong HN for silicon dioxide. The typical reaction which I wrote is SiO2 is plus 6 HF is H2SiF6 which is soluble plus water. If you have an nitride same thing process H2SiF6 plus ammonia goes away and actually ammonia does not go it actually acts like this, ammonium hydroxides okay. You balance thus this equation is unbalanced. Whenever error is kept it is unbalanced equation. It only shows forward reaction. Otherwise you make equal. Left is same as right. Number of molecules either side are equal. Polysilicon is etched. Polysilicon cannot be etched in HF because only I already started telling you that one of the test whether you are etched oxide is you add water when you find it is not sticking that means silicon has come. Hydrofluoric acid only etches silicon dioxide. It does not attack silicon but I need silicon etching. Where do you need I need silicon etching? STI. I leach STI. So I need silicon etching. At silicon etching I need something which is also of course polysilicon if you have a gate of poly you leach poly okay. So what we do is silicon can be etched in oxidized form. So what is silicon oxidation? Silicon plus some oxidant will make it SiO2 and then it will dissolve in hydrofluoric acid. HNO3 is very strong oxidizing agent or hydrofluoric acid is also if not equally good oxidizing agent. So I can convert HF plus HNO3 plus water dip the wafer in this. HNO3 will oxidize the wafer to silicon dioxide and HF will remove it. Just for the sake of it those who are still in chemistry there the HF available in the which is called fully concentrated 100% HF is essentially 49% HF. What does that mean? The most concentrated hydrofluoric acid has 15% water okay. So please remember when you calculate you should treat 49% as HF when you say it is 100% concentration okay. So calculations this ratio is from 49 and not from 100 CC if you take. So if you take 100 CC HF actually are using 49 CC HF okay. This has to be because those who are doing chemistry please remember this. This ratio is actual HF. So HF plus HNO3 this ratio is so adjusted that the rate of oxidation is matched by rate of removal. That is how the ratios are adjusted. You can also use SCL H2O for aluminium. You can also use ortho phosphoric acid S3PO4 which also uses HNO3 plus H2O2H aluminium. Aluminium is a strong metal which was used till very late. Now of course it is gone. So one need to H hydrofluoric I mean this H3A orthophosphoric acid. Of course there are if you go to the Plummer's book there is a table given for all kinds of material what kind of ratios of HF HNO all acids you must need. Whenever you dilute some kind of HF it is called buffered HF okay. Normally we add ammonium fluoride to it and we also add acetic acid. Why? Anyone? Chemistry of a man can say that. Why I add instead of water I add acetic acid. What is acetic acid? CS3CWH. This hydrogen ion concentration changes what? The pH of the solution. So when you do any reaction pH changes for a uniform H pH must be maintained. Acetic acid keep providing hydrogen so that pH is at least higher than 3 or 4 okay. So that it is acidic in nature okay. This is chemistry thinking. What to add? I must maintain pH. In etching one of the major requirement is what we call selectivity. That means I have 2 films and I am etching. I want once to be H res less and the other should be fully H. Like I put a photo resist on something and I am etching oxide. So HN which is removing oxide should not attack photo resist otherwise the whole purpose is lost, lithographic acup. So selectivity of an HN is very very crucial that 2 films it should have very different H rates okay. So this selectivity is defined by term S and R1 is the H rate in 1 and R2 in the other film. So the ratio of H rate in 1 to H rate of 2 is essentially called the selectivity. So if you want R1 to be faster than R2 then H should be larger. If R2 should be more than less than R1 then S should be less than 1. So that is how you maintain the solutions or another problem with wet etching as such and this is the important figure. I started with a mask okay maybe resist or oxide now let us say anything it could be. This was the window I opened is that correct? This this was the window I want to H okay. But when I was etching the area below I figured out that okay this is not attacked selectivity is good the resist or whatever the other film is not attacked but the lower film is getting attacked. So the solution which enters has no directionality it will etch down but it will also etch left side and right side it is called lateral etching okay. So the solution comes it etches this side it etches going down also okay. So the kind of pattern which actually you wanted maybe this was something like this. This is ideal pattern I wanted H just below that whatever the thickness this is thickness on this. But what would you do it? Sideway etching it will do okay. That means the actual pattern which I wanted to print has got extended is that clear? I want this much area now I have got this much area. So my next mask does not know that okay. I thought that this is what window you will open I will put something there now it has gone out okay. The next connection here which is here may actually touch this now but I separated that node allows me to separate by node minimum and I figured we are both are touching okay. So there is a risk involved so this additional lateral areas or lateral sides is called bias what is it called? Bias. So 1B here and 1B here so it has a bias of B 1B left side B right side and this is the film actually you wanted H2. So let us say this is SF so what mask is SM but what is SF? SM plus 2B is that correct? This is the mask this distance is SM mask H okay but what you really got the film H is SM plus 2B. So your pattern which you wanted SM has now become SM plus 2B. Now there is another problem the thickness of the film which you are etching proportionately lateral H will also increase B will also increase is that clear? So keep on etching down you also keep on increasing now. So the film thickness also matters how much B will get is that correct? Thicker film if you etch B will be much larger so your pattern will be almost not there what you want okay. However normally what we do is we will show you figure even after I reach an H and H I put this is up to 1 I will still etch over H something okay why? Well I want this as flat as possible okay this will be some circle so if I do so further you can see so I can get this area as flat as possible okay so I will do over H but if I do over H additional B will come okay. So the problem is I may get a good this transfer of image with a larger size if I really do weight etching so do whatever it is weight etching will always have higher size patterns than what mask is asking so this is major worry in weight etching. Typically to show you this this last slide okay for the day at least we define a term anisotropy okay anisotropy is defined as 1 minus H rate in lateral direction divided by H rate in vertical direction. If lateral H rate is 0 how much is F if lateral H rate is 0 as in the third figure 1 so the highest anisotropy is 1 what is the lowest anisotropy when the lateral H rate is same as the vertical H rate so it is 0 so F varies between 0 and 1 the worst is 0 best is 1 best is 1 so what is our ultimate aim to make AF as close to 1 as can you see the third figure the mask and the H is same dimensions is that correct this is SF this is SM so SM is equal to SF that is what I wanted I may play some game so that little bit bias appears so now SM plus smaller side etching so SF is not far away from SM but it is outer of this and here SF is much larger than SM so my whole trick in designing an H system is from wet etching I must try to reach this and ideally this there is no system which is anisotropically 1 what does that mean come what may some lateral etching will be performed do whatever you do even with normal ion etching you do some lateral etching will occur but our aim is that H rate should be much smaller than the H rate in vertical if I do this I will get highly anisotropic films HRH areas this is highest anisotropic one whatever is the mass dimension exactly same dimension is of the film okay is that point clear so we will like to find out next time what is the actual film thicknesses you need and what is the limit you come how close you can come because after the mass decision is decided by who I say 14 nanometer notes I am telling him this much separation you can keep now if you are separation you told designer something he has interconnect lines closer what will happen they will merge or half of both of them will go so my worries started that whatever I had told designers it must be conformal to what I should give him so I am trying to achieve AF equal to 1 how close I get I normally may try by 0.8 0.85 when I will certainly never get this is roughly what next I will do and this is not very big one last part of my course which is very important somehow I am I couldn't do anything on that but some salient features I will definitely tell and I said 50 percent of industry successes on the back end designs or back end processes we are still not finished with front ends okay so the back end technologies make all the correctness of the design you worked okay firstly interconnects all the process whatever you do is fine device how good is your interconnect is decide the success that is your back end so the back end processes failures testing is a huge area in itself okay so I will not be able to do in 30 40 minutes so I will give you some salient features failures concept you are a migration bullet go or the junction pitting there are many such issues so I will just list them to you okay so back end design read chapter 11 if you need more seriously for more details on itching chapter 10 at least read 10.5 10.5 1 10.5 to 10.5 3 where modeling has been done which I am not doing very small modeling very creep I may write also but model model for each okay so we like to show you how itchings are performed by I need models computer needs maths okay so we must create models okay.