So it's a real pleasure to introduce Lester all over Brandis today Speaker all over got his diploma in physics Technical University of Karlsruhe in one nine hundred ninety and then his Ph D. at. Zurich in one nine hundred ninety four came to Georgia Tech to work with Mark Allen as opposed. To. The as a lecturer before coming to Georgia Tech in two thousand and two where he's currently Fessor in the school of Electrical and Computer Engineering and executive director of the Institute for like trying to you know technology and when I first asked Albert to give an adequate tech lecture for this semester he said well maybe he was worried that maybe people wouldn't remember his previous one from two thousand and eight so I shook hands was there anybody here in two thousand and eight that Soledad like should. I think you're saying. OK. So I feel comfortable now so go with it so I can show the slides again from two thousand and eight is that right. I won't do that. And I will tell you thank you for coming first of all of. These days I said that before I rarely get if I rarely give research presentations I more give overview was presentations about what I was storing so I feel really weird he was looking forward to this presentation on Bryson and memes actually a field which I started working as a Ph D. student right so my Ph D. was from ninety ninety to ninety ninety four so I really started working in the resonant sensors area I ninety ninety so it's twenty five years now it's actually. Quite fortunate to stay in a research field for so long right and then to have funding over such a long period for a particular area of research this is no longer. Only thing that I'm doing in my lap but it's still a major activity of what we're doing research Weiss and as the other disclaimer that I want to do even though even though my name is on there on the slide I really must say that this is to work off the students right and they're sitting right in the back of the way in the back of the room right than I have later on a picture and I have an acknowledgement slide with them as we all know this would not have happened right so it's really to work off the students especially since I have to. Administrative appointment right even more low on the students and then another thank you goes to the staff. That facilitate all that work and a lot of the staff members also here in the room right and maybe later on at the end when I showed the acknowledgment slide we can all stand we can all stand up. Having said that what I really want to do is I want to talk a little bit about in general why resident so mems micro electro mechanical systems and then really show two examples one is a magnetic field sense sort of is a nice example of what I call the systems level approach to resin and memes and then the work on chemical sensors where we have probably most experience on the on the resin and sensor side which is what I call a device level example for us and names and so having said that what do you do when you resonant mamzer when you use resonance sensor said sometimes I say resonant ma'am Sometimes I say resonance sensors usually for me I mean the same right so I mostly talk about sensing applications off memes and using resonators for sensing up to cations often. Memes So what do we do we explore the residence to mechanical isn't Anscar Rock the roost stick over my comic nickel structure and this is actually a transfer characteristic of a cantilever beam out of my Ph D. thesis so I measured probably something like nine hundred ninety two or so my but it really shows the point is the amplitude transfer characteristic right on the top with a resonance peak and then it shows the transfer of the face with a hundred eighty degree face shift across to resonance and what you do for resonance sensors is quite simple you expose one of the resonance characteristics of a micro mechanical structures to do sensing light and most of the time it's actually the resonance frequency that we explore right. But you can also explore the quality factor the vibration amplitude to write or read the face relationship for sensing applications the examples that I will show today really explore the resonance frequency the. Quantity did we measure as a function of the whatever we want to measure to measure and Alright so what do you need to do for a resonance sensor that works. By exploring the resonance frequency as. The sensing quantity is we need to measure the resonance frequency as accurately as possible and I will show you in a moment how we actually do it out right and so did the thing it's actually pretty Ressa tile I do if you scan the literature. You will find almost everything measured using a resonance sensing principle right. Most of the time it's kind of a two step conversion right so you have a first signal conversion where you look at what you want to measure and you transfer that. And into an intermediate signal which directly affects a cut the wrist tick off the off the resonance behavior of the Michael mechanical structure and that might be a damned thing that might be a driving force it might be a force in general General apply it to the residence structure of metrical changes mass changes a material property changes you name it right. Sometimes you can do it directly you need only one conversion for example if I want to build a first sentence or a kind applied to force directly to a micro mechanical structure and that first can change to resonance or up the wrist tick off that structure right so this would be for me kind of a run step signal transduction in most cases however it's a two step it's a two step approach and again if you scan the literature almost any Everything has been demonstrated. To be able to measure using a resonant principle so why wasn't sensors again the first thing I said already my pretty versatile sense of sensing concept if you look at the underlying structures they are very simple they're actually boringly simple right so a can to live or a can to live a beam to fabricate is very very straightforward to do actually you can do it as part of a class right in the Mechanical Engineering six to two nine class I think they do fabricate a can deliver us part of a spot of the slap cores and. You could make a resident sensor out of that very very easily so for me very very simple structures and very very well understood structures I will show a differential equation in a moment I'm a physicist I left differential equations so I have to show some equations right there very simple and again the structure is a very very well understood you have already with the when you used a frequency as the output signal almost a digital signal right so simply following whatever whatever. Signal processing circuitry you need at the back end and in the end frequencies can be measured quite precisely if you can count long enough right so well that makes. Resident sensors I think a very very interesting approach to sensing So I promise you I'll show you a differential equation. To make my point. Of distinguishing between device level frequency sensors and system that frequency sensors so round every individual resonance we can describe the resonant behavior of a micro mechanical structure by a simple spring mass stache point system or a forty electrical engineers maybe now see resonant circuit right. When you look at the first spell and right you have pretty inertial force mass times acceleration. The damping for stamping core fission times of a loss of T. and a spring force spring constant times deflection that equal whatever first as you apply externally to destruction are all right and. It might not be straightforward to find the M.B. and the K. but every. Resonant peak of a micro mechanical structure can be described by that by such a second order system all right now the way to operate these devices is the following will apply an excitation force that exactly can sort of the damping force right and this way we sustain for example a Bedouin resonator into an amplifier in feedback loop the amplification part of the driving force that we apply to the system basically cancels out that them in force because that's the last for the last term in the differential equation right so I read right. In for a sustained a vibration is essentially the damn thing forced to cancel out the damn thing force and without the resonance frequency becomes proportional to the Screwed of the ratio of spring constant and mass and I should say effective spring constant and effective mass for that particular resonant moped and if we have a right to measure and affecting K. or M. we have a sensor right and its simplest way for our chemical sensors. Measure what we do chemical said we want to measure observe it to the surface of the cantilever change predominantly the mass of the cantilever and we can send start quite easily right and the way we do that is on the next slide right we take our memes resonator and we embed it into an amplifying feedback loop right and we might have to do some face adjustment to get here a signal that is really proportional to develop. Off the off the transducer we feed it back and then we have kind of a self assimilating device that exactly resonates at its resonance frequency desired resonance frequency if your I mean changes that frequency changes and we can try to the frequency which is then a function of what we want to measure right that's the simple way. Another way to do it going back to the previous slide is this what I call system frequency sensor so in here the idea is that when you look at the first listed you apply it to your resonators Yes you have again that F. one which basically. Cancels out the damping floor so that you can sustain a vibration in the amplifying feedback loop but if you can apply an additional force that is on. To. The deflection off your resonator we are to do. To the absolute ration of your resonator if you go through the mass and it's really not not difficult to go through the mass you receive that then the resonance frequency becomes a factor becomes a function F. boosts proportionality factors here K. at zero M. at right so K. add an M. at right and now or if you find a way of making M. at a function of what you want to measure you have what I call a frequency level freq system level frequency sensor rather than a device level frequency sensor what the advantage of that is I will show you in a moment Well I tell you right now this is something that you can potentially turn off and turn on at your will so you have an intrinsic differential measurement capability in the system you can make a measurement without the impact of the measurement right by turning that term off and then you can have a measurement of frequency measurement read the measurement move the measure run affecting the resonance behavior by turning this time on and I will show you in a moment how you can actually how you can actually do it right so in a in a system schematic this looks this looks that way right so we have our primary feedback loop. Which really ensures that the resonator is continuously vibrating at its resonance frequency. And then we'll have a secondary loop. Or a modulation loop right and so here in this case the signal is not to the velocity but there is a phase shift which makes sure that is proportional to differ. Miten if this is a function is a function of what we want to measure. In this loop just with a simple switch on and off and have an intrinsic differential measurement capability right there are some seats here in the front if you if you if you want to sit I mean there is at least for seats here in the front please feel free to come forward and said. Everyone but that's pretty straight forward so how can you do that for the device and so this is this is a work that actually goes back to the first student that worked with me here at Georgia Tech and we probably should this already in we probably just already in two thousand and six he was actually a student at the TH and he couldn't fabricate his device as a D.T.H. So he came for a single one or two years in the beginning right so that's. It's the student's name that's his work so that we do is we use a kind to lever we to use a cantilever resonator right and you will find this as a seam throughout the resonance and so work from my group family excitation using resistive except ational amende and P.'s are resistive detection for this primary loop to sustain the vibration off the resonator So the only thing that we have really done to make a magnetic field sensor in this case is we have to design the current loop around the perimeter off the device right and so that current loop will feed with a current that is proportional to the two deflection of the cantilever right so we take the signal from our P.S. A resistor goes through a circuit to make sure that we apply a current which is then of course an AC current because the can to lever is why braiding. That is proposed to deflection off the cantilever all right and then there was something that the physicists know was learns first right so in the presence of a magnetic field if you if you have elections moving through this through duce conductor day experience they experience and the ones first that is normal to the direction where the electrons are moving into the direction of the magnetic field so if you say the electrons are moving this way here the magnetic field is in the length I reckon it Jane generates an out of plane and out of plane from there so not out of plane for worse we generate an additional driving force acting on the resonator that is proposed to a deflection off the resonator so it would cost at frequency as shown on the previous on the previous light and it's here again right so the resonance frequency now becomes a function of the spring constant of the resonator of the mass of the resonator or effective spring constant and effective mass and the spring constant is food a module aided by this addition. Force term which is a function of the applied field right. Producible doesn't work of course it works right so this is a device that was fabricated this was actually not fabricated at the Georgia Tech clean room this was done in a commercial CMOS process you see the cantilevered beam is a little bit more an odd structure off the cantilever beam and you see that cone loop here right and a closed loop is not closed along this line but to reduce crosstalk we close stood on the round the round the my commish this is a micro machine cavity to release to release to can't deliver and it's not only one current loop but it's one two three four and I actually really think it was eight loops. With the metal is ation off to see Most process at those space right we were right in but here we widened it's here at the tip of the at the tip of the cantilever. To Half Moon length that generates Lawrence force to basically increase to increase the Lorentz force and again this was this was published already in two thousand and six in the Journal of memes and a student. He's now back in Switzerland. Working in industry no longer. Does it work well it does actually work quite amazingly well. So this is the frequency shift off the resonator as a function of the implied magnetic induction from zero to ten merely Tesla right and if you clicked in there the next one some data comes in the base resonance frequency of this device is one hundred seventy six killer Hertz to Q factor in air which gives us the reason Lou the frequency resolution of the system is about six hundred this gave us at that time a frequency stability of about twenty five million Hertz So that's the frequency change that we can result in if based on the frequency noise suddenly measured if you look at the sensitivity to slope of this curve it's sixty cooler it's protest law and if you extrapolate. A limit of detection using this short term frequency stability and sensitivity you find about one micro Tesla so you can reserve the Earth's magnetic field using the systems without the need of any magnetic material why does this pure silicon technology. You can do with this device only in one direction and you could do a T.D. sensor however simply by rotating the design by ninety degrees right three to you sensor is a little bit is a little bit more difficult but again no magnetic material or anything like that. And. One of the things that I like particularly in transit drift compensation right so you turn off that secondary feedback loop and you have a reference frequency did you measure you can immediately turn it on again right and you can do essentially in time a differential measurement that gets rid of all the drift. But there's a slide to the non-linearity in there which is obvious right because if you go back to our Because if you go back to our physics it's not a linear relationship or frequency change in the magnetic field but it's here square root relationships that will lead you right for small magnetic fields and if you look at going forward if you look at the non-linear T.V. right so if you subtract the measurement from the linear fit this is actually this is actually the non-linearity did you find in terms of in terms of hertz so there is an intrinsic non-linearity in the system but that's not surprising for this particular thing so. This here I find very intriguing right so there is an intrinsic sensor offset compensation by switching off this additional loop I'm still looking for that for the chemical sensors right so if we go back to the equation. Challenge to the Ph D. students here. If if in this case I did escape or do we could make a function off a chemical measure and we could have a sensor where we have an intrinsic offset or drift cancellation which is one of the big challenges in chemical sensors right so. I haven't found a solution but I'm sure one of the next students will find out. And solution to it all right so this is to work on the. Magnetic field sensor and so this now brings me to snow brings me to the chemical sensor and I have about twenty five more minutes or so is that right so we're actually not doing about. Chemical sensors so. Let's put all that in here so when I talk about a chemical sensor it usually looks the way that we have a transducer in our case this is a resonator we have a sensing film that we put on top of our resonator that sensing food celebs. D.D. and Allied of interest right and other stuff as well point this comes to one of the big challenges. And in that sense inlay is. Kind of. A transduction from the chemical domain to the physical domain happens in the sense that. The light in the sensing film course has some sort of physical material properties change off that sensing film which we then pick up with our transducer oftentimes we have signal processing at the back end in order to get some sort of standardized output signal right and here we try to integrate essentially on chip in addition there is a lot of things that you can do on a sample preparation side you can filter you can do pre concentration we have some activities in that area as well but I will not talk about that today but really concentrate on this on this sensing piece here. So the transducer if you look at the transducer here I think there's many many rays to build that transducer and there's a lot of literature out of that I mean actually we have Jenna. Here in the here in the audience he is probably the godfather of chemical sensors. In the room fly clear righty has written textbooks on that and I think the division into electrochemical Sensis mass sensitive sensor so my son's an optical sensors goes back to you right. I cited Isiah to Andreas him a man who knows very well this is an overview article on miniature Rice chemical sensors that we did we published in the Proceedings of the AI Tripoli you see also already a couple of years back in two thousand and three Just when I came to just when I came to Georgia Tech so you can divide the transducers into these four categories right with electro chemical sensors being still I think I followed the largest off the categories. For the five percent and. Areas that I highlighted in here Georgia Tech is areas where we have currently research going on but what I will concentrate is on these mass sensitive sensors and I were framed from saying this is the best you can do you can focus on right because whoever you ask will give you probably will give you probably a different answer here so this is what I will focus here and I know the sensing this is also really really important this is another publication out of our mutual friend from a mutual friend of Dale's Hill a month where he looked at different sensing films and there is really really a lot out there and as an electrical engineer I must say depressed well maybe as a physicist I must say the best thing is to collaborate with a chemist on the chemical on a sensing film site because there is such a wide were right reviewed. Of course on polymers sensing films poorly Psylocke saying poorly you are saying yes and you see some of the sensing that can be done using dad we primarily look at one time log on it compounds but with an air samples but also in liquid samples but there's a whole lot that you can do in this area as well we have played a little bit around with bio sensing but that say I tell you this is not an easy field and again you better leave it to a good collaboration with either a chemist or a biomedical engineer right in order in order to get that in order to get that going. So why resin and chemical sensors right so. Again I think it's a big very simple sensing principle and what I really like is I know what I measure or think I know what I measure right I really measure a mass change some of it some of these other sensors is often harder to find what you are actually measuring right but because the physics is so simple for these resonance sensors I rudely know that I should say that my students measure a mass change all right that they were doing this isn't this is a picture of the cantilever cantilever resonators on a on a single chip we would use we would. Cantilever store however the structure would look like with our sensing food polymer right and we look at the change in the resonance frequency decreases as the sensing film absorbs and the light from the from the surrounding you can do that was really a sense of principle as we focus on these candle of a type structures we really look at the environment for money towing. But there are points of people exploring this in an area of medical diagnostics breath analysis for example. And one of the nice thing about microfiber creation is if you can do one you can too ten you're going to a thousand you can do a million right so you can make sense arrays and the sensor array can help you with one of the greatest challenges for these chemical sensors which is a selective A T. challenge to distinguish between different an alliance and has been a lot of work in that area but I want to show you a different approach to it out to do some activity challenge. This is how resonators look like they can be simple cantilever type structures they can be more complicated we call them describes nadirs that kind of these kind of vibrate kind of rotate in the in the plane we always use in my lab some Aleck's citation and P. is a resistive sensing. This is some equations again to limit after Texan us one of the key parameters is limited on one side by to frequency noise of your resonator dots at Delta min and then by the sensitivity of the resonator itself so we should want to improve the limit F. the Texan you make sure that your frequency noise degrees is and you generally do that by increasing the Q. factor off the resonance you make sure that your sensitivity increases for example by putting more sensing film material onto your onto your kind a lever structure. Designed on fabrication a pretty straight forward. Again. We use to resistors we actually use only one to excite these resonators into resonance we just apply a C. signal to use heating resistor superimposed on a D.C. signal but that's all. Detail in order to avoid frequency doubling. That drives the resonators in the desired in the desired resonance frequency and then we sensed a vibration with. P.S. a resistor for resistors again silicon resistors that are ranged in a week so I'm bridge and we designed the reach so I'm British smart so good at it it's census the mode of libration that we are interested in and not the other modes of white Gratian because every mechanical structure has more than one mode of fiber Asian. Fabrication is pretty straightforward. Patrick I think it's what a six mask process. Six seven mouse process of course is not a single mask process and in. It is not one hundred percent but it's a fabrication process. That we have been using for the past I'm yours since two thousand and three we have been using it essentially since two thousand and three and it's it's adapted from a CMOS fabrication process that we used back when I was a wreck so there is really no not much magic to it we use both micro machining at the end initially we use Kalay churching from the back of the way for now more and more we use deep deep reactive violent I-Ching to basically release the cantilevers in a back side step and then we have to fun side step how we really define the geometry of our resonant structures. What are the challenges right so it should be an educational seminar and in reality not all is good I think right so if you think about resonant chemical senses but I truly think this is for I make zero sense as in general the grand challenges are sensitivity to three S. here sensitivity selective A T and stability right and I will really not talk about. At how we address this. Stability issue but we actually have done work where we use this system level. Delegated I showed you at the very really beginning to actually do a drift compensation it's not as easy. As we thought it was in the first in the first place but you can you can do it and if you're interested I can point you to some proper cations I will really talk about these two but on the sensitivity side we have to do is reduce noise right reduce frequency noise and we do that by exploring. Resonance modes that have a relatively high Q. factor in the environment I am not interested in vacuum right because chemical sensing in vacuum is pretty difficult so I'm interested in either air or in a liquid environment and then we have played some games being able to increase the sensing film Well you without affecting the cup the wrist itself the resonance these are polymers said we put down and you can imagine a pile of lossy material if you put it down on a cantilever You will affect Q So just putting it down on the cantilever and hoping more and more polymer and hoping that the sensitivity or not at the limit of detection will improve. It's not going to happen I know it's going to happen because you kill with more and more polymer actually Noyce correct the wrist tick so if you're resonator so I will show you a little bit what we do in that case but I'm really excited about recent work that we have done in this in this area to Dish and only use higher order sensing in order to get some selective A-T. either with the rays of the same kind of device that are coated with different sensing films or with different kind of measurement principles in a rate we. Right to look into is to analyze trends in signals for improvement of selective A-T. and we're not the first one to do that in general in the field of chemical sensors but I think we're the first one who we really explain the road out in a resonance and sing in a resonance sensing environment so. How do the American say with a grain of salt salt right so there's a scale kind of my my little joke here. Can factor we have to improve Q. factor right when I started when I started this work back in the early one nine hundred ninety S. as a Ph D. student you see I was obviously not a good engineer my cue factor and it was ridiculously low two hundred and then there were a couple more Ph D. students over time working on it and at some point I kind of blotted the pew factor that we measured in air over time right and I found an exponential relationship and I thought here well it's not Moore's Law right but it's maybe Brandt law but I can deliver a few factors never only wait long enough I have to sensitivity that I want to write. Well history sometimes tells you different stories. This is how this plays out. This is how this plot continuous over time and really rude to Rumi stuck around this about five thousand or so west a cool factor I must say this is an era right this is not in vacuum or anything like that. And you could I mean you could draw a couple of wrong conclusions right so one wrong conclusion would be well in two thousand and three he moved from E.T.H. to Georgia Tech right. But truly truly truly. If you're limited by by the physics right and over the years even then even though we have. Unlimited around five thousand in the Q. factor we learned a lot why we are limited. And so this is an example this is not for the more exotic structures but for priests magic cantilevers so just normal diving board structures we've looked at you see it is there is theory behind it if you would perhaps to cue factor that you measure in air as a function of this ratio here length to thickness and a sickness to a power off one point five. And we have measured I don't know how this student can know snaily. I think he's happy to not measure kinds of levers anymore. He has measured hundred so often characterized hundreds of them and if prodded them all interests in this graph here and you see there's a distinct peak white so you can do it really well or you can do it really wrong aren't when I was a Ph D. student I was probably down here right so really really wrong. But if you design a dimensions of these kind to live first right you can be at a smart spot in between where you are suppressed most limited and in between where you are in a Demp ing limited right and even with a simple cantilever structure you can increase the Q. factor too as I would say a thousand to fifteen hundred range in air right I know for the vacuum guys this is nothing when you talk about vacuum but this is really in air. So we learned about about it we also worked with a lot of these devices in a liquid environment in a letter environment and also there I had collaboration is longstanding collaboration's was two professors Fabian Yasi and Stephen Heinrich at the University of Marquette and. Created the theory behind that even in liquid if you do would run. Eight You can achieve Q. factors let's say between fifty and one hundred simple person Matic cantilevers if you make more complex structures you can even increase to a couple of hundred and there is a professor at. U.T. Dallas who has shown us think two three hundred values of Q. factors in what environment. The message is can affect as you have to make can deliver short you have to make them white and you have to make them stick I could tell you the physics behind but in the interest of the slides that I want to show I believe that I leave it out to two questions so this is I think my group's bridge resonator at the moment do you see how it's vibrating. So it's basically kind of like a half disk that is rotating up and down right so we're using implying resonance modes rather than out of plain resonant modes because they're much less. Media Right they slice essentially through the air or through the liquid rather then pushing it up and down like a diving board right so we're cool factors are much improved by using these in plain resin and modes this I don't know how to show it again. But maybe have to go back. And hear so now you see to vibrate this to right. This was actually measured in France using the stripper scopic system that can extract the characteristics of the device we draw we drove it pretty hard with D.C. and forward so you see the vibration amplitude of this particular device at eight hundred kilo Hertz is about one point five microns so that you can visually the CD The vibration amplitude Dick factor that we extracted from the transfer characteristic is about five thought. It was an air. This is our favorite device. This is a cantilevered we drove pretty hard so long cantilever rides are just a snow white grating at twelve micro meter at one hundred fifty two hundred just a Q. factor in the air and you see dimensions of the cantilever this device we actually could also observed a second to the second. Basically which has one more note aligned along the length of the longer length of the cantilever and then again we do is we use cantilever square to sensing film. These are two of our. Most you sensing films again we don't do too much research at at the moment in exploring new sensor films P.-I be. Chloral hydrate is two of the most often used ones in my group the way we the way we coat the cantilevers is either by spray coating you see this here right so everything is essentially coated we have combined that with the shadow and then you can get some sort of localized deposition if you combine disparate coding with the shadow mask we have this a played around with. Coding and this is a result of it so what we've done actually in this case we at recess recorded the bass top in. The. Head part of the resonator and that contains the jet at the Jet A. Polymer solution and this way in this particular case we've got quite nice localized position you're using in chatting kind of a disclaimer here it's not as trivial as it looks like because. When you use the printer in the clean room you will limit. That was the sort of them said you can use and they might be incompatible with the polymers said you were using. The viscosity did you get off the resulting solution might be incomputable with with a range of Wisconsin the. System can tolerate So there's a lot of up to my zation work that has to go on until you get a nice result like that initially with sputtering I could also show you to the horror pictures instead of the nice picture that's the nice and results typical film thickness is are in the micro meter range that we're using one thing that I want to say here in addition is president of the Palmer in an area which is not undergoing strain right when you look at word has to count to live a bend it primarily bends here close to the edge so when we look at how our how much just a polymer now affecting to Q factor by doing that by doing. In the head region only we avoid this drop in Q factor due to uniform the Brazil the poly marriage material that we did initially. So that helps us with the best prop we actually weigh so we remove silicon which is essentially nonsense and material and we play said with polymer which is sensing material and we also encourage improve the sensitivity to sway and you can let the students go crazy right I mean they played around with the. Recesses and printing this was work done by Chris Karen who just graduated recently you can group nanostructures on top of the on top of the cantilevered carbon nanotube type structures that we have grown out of in a low temperature process. We have in the markets. In the Marcus clean room and you can now functional Rice nice. Not you using sensing films we'll try to do that with peptides to do to do sensing bows in air and liquid environment and then publish some preliminary work in this area memes in two thousand and eleven this is work done by Beardsley. And so story this is the next slide this is a typical measurement that we do about five minutes is that right. Five to seven minutes OK this is a typical measurement that we do. We have a customized set up where we can expose our sensors to find concentrations of an ally so you see here but D.N.I. concentrations are fairly high in the thousands or hundreds to thousands of P.P.M. in this case this is a combination of we expose them to tall you in very poor in air and the resonators coated with about one and a half to two microns of the right and then we have a valve which exposes the cantilever to the to the an allied concentration in a flow through cell and you see that Rep It changes the frequency because of the exposure to in this case one thousand ppm off and the light and then you switch to reference gas again the signal goes back really quickly you doubly on a large concentration double the signal and so on it's very linear. It's very reproducible right ramping up ramping down. It's just a matter of off in a matter of a couple of seconds or tens of seconds need a frequency changes to analyze it is absorbed into the into the sensing film sensitivity we cannot measure Well actually we can now in the mean time but we haven't set it up yet we brought a system that can generate much lower concentrations. In the states we looked at what's the sensitivity. The sensor What is the frequency noise so does this is in this initial region right and you can extract using the Allen variance Mesut the frequency noise of the resonator and with that we can extract we can extrapolate the limit of detection which for this combination is of the order of one P.P.M. However this is extrapolated from measurements at a thousand ppm right and really not done at low concentration measurements and I did something that has to be followed up but this is the most common standard in the literature do you wish the sensitivity and the noise in order to extrapolate the limit of detection is also works in liquid. I do not want to focus on that because I really want to show you this idea of the transients now. So the challenge challenge that we have here is that when you measure one hundred that. That's a one hundred fifty Hertz frequency change you don't know if this is because of fifteen hundred ppm of total you or I say a number of thousand ppm buttons ino or two thousand ppm from sure if I'm doing the right way you're all. Right you cannot have selective A T intrinsically with this one sensor so a couple of years ago we did measurements were we looked into these transients right when you when you experiment with the sensing film to be on a light right and so this is normalized transience normalized frequency change as a function of time when you are exposed to an alcohol serious methanol or less on a live Soprano right and you see the diffusion times are not surprisingly distinctively different between D'Anna lights just because the bigger molecule is just take longer to diffuse into the sensing film No no magic against behind in those the. Case load we needed to do what we needed to do is we needed a complex gas mixing system to generate these transitions right and if you want to put that into a into a smartphone or into into a device which you want to use in the field it would be much much nicer if you can generate these trends Ians on the device itself right not needing a big system with mass flow controllers and valves to do that right. And so a little bit thinking going behind that there's actually physics behind disruption of the end I read into the sensing film we've just recently learned that you can simulate it perfectly well using finite element models for example. And you can extract true data you can extract the fusion. And quotation quite fissions from the state but again I spared the details of that and really want to show you this so did idea came up to use the first dimension and in this case the first I mention is not time but it's temperature right another T. because the partition core fish and so the amount that kind of tells you how much an allied with diffuse into sensing film it's highly temperature dependent right to get something that the chemists know for a long long time so that was cantilevers we can drive out the I know light from the sensing film and if we could move them now rapidly we could look at room temperature essentially how do you analyze this diffusing back into the sensing film right. So we magnify the kind of liver design a little bit and we means Chris Karen my student right. Now the head the region has two resistors with which we can heat the cunt a liver and about the hundred million rats so this is the. Power you see but we will see later we don't have to eat them only very very shortly he'd stand up by seventy one degree and the rule of thumb placed at the partition quote fission changes by a factor of two about every ten degrees or so as a rule of some. Reserve a degree we drive out most of the end of light and a good thing else so you remember the diffusion times are a few orders of seconds or tens of seconds if you look at the time constant off the cantilever it's in the order of milliseconds right so you can heat and cool it much faster than the an ally just a few zing in an out of out of out of sensing film and so this helps them to do a measurement like that right so. This was done in our see in the set up initially we experienced a cantilever using the set up to carrier gas and then rule introduce in this case I know I told you in the frequency changes right so this is Ralph generated basically a signal trends in that is generated by the fluidic set up but then we don't do any switching anymore of the analyzed concentration we keep it constant and the only thing is that we pay you adequately the cantilever. And I really explain what's going on here in this in this in the strongest case for the mirror what's a feeding problem where you see that the frequency initially immediately goes down this is because you heat up the candle of a very quickly just changes the mechanical correct of a stick to young smart you know something can deliver and the frequency goes quickly down right in the matter of milliseconds and then the red part is and the light being pushed out of the sensor right because it's a hit it can deliver not a partition quite efficient is reduced so you have to drive out and the light from the cantilever and then you turn off the heating the frequency rapidly jumps back. Because the can deliver rap. Roots the young smog changes and cause that increase in frequency and then at room temperature essentially the I lied diffuses back into the sensing film right now if you look at this signal transition and this transition and and you prep them in the right way they are identical so deep information that you can extract from this family generated trends and it is identical to the information that you generate from the realm of switching generated trends and I really have to stop now I guess right so we're playing around more with we can do now measurements so first with the help of Patrick another Ph D. student we can now much much do more measurements of the frequency Pro time so we got our trends and signals a lot more accurate we do measurements with an allied with out on a light and then look at the differential signal and you can learn a lot from that this is really ongoing work initially we thought we have to heat for lambda's this three seconds or so eating more recently we see it's not needed we do you only need to heat for a couple of milliseconds essentially driving down the average power consumption a nice device is tremendously right so this is really ongoing work that Patrick who is in the back of the room is continuing to wrap up. I couldn't have done this without the help of the students. Right and I see that isn't the this is the cantilever students this is where the cantilever students over the years. Most results that I showed were from Lou Beardsley Chris Karen and now Patrick gets Patrick is in the back second to back room if you have any difficult question he will answer it. And I really couldn't have done it we couldn't have thought it over the years with the help of the staff and I think I want to have them stand. Right I think I'm the first speaker. In a staff to stand up but I want to do that can you stand up my staff police everyone out now. This is some of the guys. Who made this happen in the clean room right so I think these devices are not a lot but I showed over the semester here couldn't be done without them I really really want to emphasize that there was a number of faculty members that I have collaborated and I've been fortunate to have a serious off and S.F. works that have basically funded that have basically funded this research we've published a book. That came out in a prim. Surprise surprise resident memes. She asked me would you do it again I would say you know it's a lot of work that goes into that but I felt my footing only I think one and a half chapters or so right so it's really a lot done by. The chapter to chapter or first. That's my group in the Marcus building so just to give people quests he graduated recently is now with Harrison Florida. She was doing work on the circuit site he is working with T.I. that was Patrick he is the lonely survivor bear right from this not from this picture but from this kind to live a group and he's carrying on he's carrying on with that with that word. Spirit and chums who know also in the back in the back of the room listening to the talking what I have to say so without I close I'm already two minutes over. Thank you for your attention and I really appreciate it. Yes. No diff is changing it's. We could we could be another way of doing the feedback right where we adjust to current right basically did escape be constant so that it is we could do that we have never tried that. So can we. Just. Leave it uncovered. That might very very well be that might very very well be maybe we can talk we can talk more about. OK Any other questions. Yeah. How do I define it so you basically take. You take a measurement for it if you want to maybe if you want to measure short term frequency stability at an average time of one second you take a measurement every second you compare you compare neighboring measurements so differential of neighboring measurements and you average over that differential stick Alan variance method it's very very well known and as you do that for different times right you can see and so do you as a group is probably the best group to talk to Georgia Tech about that because they do that for their resonators all the time you pending. And then you get kind of a corrupter mistake where you have frequency stability as a function of the average in time moment and you can read a lot out of that you for example can distinguish what is my drift kick in worse is worse the right noise contribution worse the flicker noise contribution and so on. Thank you.