let's get started it is my pleasure to introduce beta trules clava uh she is a senior superconductor IC designer at dawave and is part of the processor development team where she designs circuits and draws layouts used to fabricate the processor prior to this role beta was an analog IC engineer at on semiconductors and she has a master in electrical engineering from the University of tenta in the Netherlands beta we are really excited to have you here please take it away with your talk on the basics and the hardware of quantum and needers thank you so much Marlo thanks for the introduction and thank you the organization for having having me here giving this talk very excited so yeah my name is Berto I'm an IC super conducting engineer here at dwa and today we're going to talk about some basics of quantum analing and also how we implement it here in the wave but first i'm just going to give a brief introduction on what uh we doing here in rewave so D we've been around for 25 years and in that time it released five different generations of unhealing quantum computers we provide access to everyone to do uh these computers through our cloud service called lip uh so lip is a plat platform that dawave uh created developed and maintained and we also offer developer tools and Professional Services so to help people interact with our processor my presentation today is going to be mainly about the hardware part but on Friday my colleague Sara is going to give a software bootom and she's going to give you a tour through like the uh lip platform and how you can run uh some problems in our processor my OB objective today will be that you understand the hardware enough that you can have it an idea of what's happening when you run a problem on Friday I will be successful if somebody can imagine what's happening there behind now uh just a brief overview of why I think analing is interesting not just for research but also for businesses like our Quantum uh or an Quantum analing processor solves optimiz ation problems and luckily for us optimization problems are everywhere so from Logistics manufacturing to machine learning we already have uh many customers who are using our processor to solve their real world problems I really like an example which is like Save On Foods I know there's people from everywhere in the world here but uh that's a big grocery store here in Canada and like I go there shop every now and then so we are working with them to automate the scheduling of their food track deliveries I like that example so maybe we can start with my presentation today we're going to first do a brief introduction of what what is quantum anink and then we're going to look at how we implement it here in thewave luckily we're going to have 10 to 15 minutes for some questions after that so let's start what's quantum analing so Quantum analing it's a process that uses quantum mechanical effects to find a global of a a global minimum of a function now in optimization problems you know we have this many possible combinations of variables and constraints and we try to find the best combination to do that we defined an objective function and then we try to minimize that objective function so you can see there's some parallelism there we have our processor has a net of cubits and couplers that we're going to talk about it later and what we want to do is we map a global function um this objective function of our problem into our processor and we let Quantum an link find the global minimum of that uh function how do we do that first I'm going to introduce uh one concept which is like the system hamiltonian system hamiltonian is a function that Maps states of a system to its energies let's think about an example very easy one we have a system which is a ball and a hill and then we have two states the ball is under the hill the ball is on the top of the hill those are our two states they're going to have different energy hamiltonian is going to as like ball on top of the hill has this higher energy ball on top on bottom of the hill has this lower energy now we also sometimes use uh some other concept which is like the ground state of a system ground state of the system is the lowest energy state of that system so Quantum veling is going to use What's called the adiabatic theorem the adiabatic theorem the tell us that if we start from the with a system that it's at the lowest energy state of the hamiltonian and we evolve this hamiltonian slowly enough in our and final hamiltonian we are going to be at the lowest energy uh state of that new hamiltonian now what I want you to remember is that we will end up in this lowest energy State we did map or optimization a problem to our processor so after running our Quantum anink if it's in the lowest energy State we have the solution of our optimization system let's take a look let's try to depict it like to have some cartoons that can give us an image of what's happening so we start with our initial hamiltonian and we prepare we will prepare this uh our system so it's in the ground state of this h neonian now physicist like to talk about wave function so the wave function you can see it in blue in this cartoon it gives you the probability of the system to be in a certain state in this case it is localized Above This Global minimum then we run our anal slowly we try to uh we start incorporating our problem our like bringing the hamiltonian more to or problem and you can see that the landscape the amount of possible solutions it starts to sweep it starts to move and our wave function delocalize it uh goes over all the possible possible States this allows tunneling I like to imagine that as a wall in classical physics the ball would just roll and if it doesn't have enough kinetic energy it would not pass and go to the next Valley we have in this landscape but because the wave function has delocalized and it's going all over this Minima the B can tunnel through this uh barrier and we can end up at the global Minima of our final hamiltonan so now we're going to take a look of the hamiltonian of our processor our quantum processor it's going to have be described by this hamiltonian that we see at the top of the page this consists of two parts we first term here we call it the transfer field hamiltonian and we prepare it so it's in the um it has a single Minima and we know that we are at the lowest state of this hamiltonian the second term it's the problem what we call the problem hamiltonian and in this hamiltonian will already or it will have uh or cubid biases or coupling biases it will have our problem in it problem information now we use these two variables A and B to do round the anal so at the very beginning a will be at the maximum and B will be at zero so we have or transver field hamiltonian as our initial hamiltonian now during the anal we're going to sweep these two weights weight parameters until we completely eliminate eliminate our transverse field Hiltonia and we are all in this second term and if we do it slow enough our uh final hamiltonian problem hamiltonian we will be in the lowest energy state of that hamiltonian which will give us the optimal solution solution let's take a deeper look at this problem hamiltonian so as I said this problem hamiltonian has information about our problem in here we have the hes the hes will be the BS on the Cubit this uh biases are going to are going to tell the Cubit if it needs to favor one State or the other so if our cubits have a zero one state you could tell it maybe just go towards the one at the end of the nail the JS are going to tell us or tell the cubits and some coupled cubits if they should go towards the same direction or different opposite direction I like to think uh I like to put an example on this so imagine uh we're trying to solve a problem where we're trying to seat people in an office there's people who really likes the window sits next to the window there's people who do not like the seats next to the window H is going to tell us this person prefers this window seat go towards one or uh this person does not like the the window seat light reflects on their monitor go towards zero ages will kind will depict the relationships of the people two people can sit together in those two seats because they work well together let's bring both of the states of this cubits towards the same direction yes or no or up and down so all these H and JS which give information about our problem are going to be programmed on chip through our controlled circuitry which we're going to go through later after I will give a bit of more high level introduction to the hardware so I'm going to do a bit of big jump here and I'm going to start talking about our Hardware but we're going to come back to this hamiltonians and I hope we can bring all these Concepts together so how does process our processor look let's start by very top level here on the right there's a b big box it's quite big like usually person will stand above between a and v average high high person this box contains our processor but it has multiple user uh uses for example there's the electronics that all it allows us to talk to our processor it also works as a electromagnetic noise Shield uh if we would have electromagnetic noise it would interfere with the computations of our processor and it also cools down our processor to ultra cold temperatures now I'm not here to talk about any of these uh hardware parts I'm here to talk about the B tip the processor here we're going to find this microchip and there's where all of our cubits are so the uh processor uses super conducting cubits uh or cubits are made from Stand semiconductor fabrication uh processes and so it means we have this multi-layer stack with similar metal layers which is the ones we can see here in a light so this is an example crosssection that we could see if we just cut across this microchip here and in the darker darker areas it's on dialectric so we can build our cubits by using all these metal layers and now to make them super conducting we need to cool down these Metals below the critical temperature of the metal that we're using but we're not stopping there we would also like to minimize the thermal noise which our cubits are sensitive to so we go way below that critic critical temperature so let's talk a little bit more about super conductivity now as I mentioned uh we need to cool them down uh we need to cool our processor down so our uh cubits become superconductor what happens when they become superconductors superconductivity is a state let's think how a metal has some proper physical properties when it's solid but it changes it properties when it transitions into liquid if you cool some metals enough they will transition into super super conductivity and they will show these two properties the first one is given away by the name it will have the metal will have zero DC resistance which means if we have a wire of super conductor and we send some current down that wire we're not going to see any voltage drop the second property is flux expulsion usually a metal will uh we put a metal in the middle of a magnetic field magnetic field which just penetrate it will go through this metal now if we have a superconductor the this in know in the broad term it will just expel this flux it will just go around it so thanks to these two properties we can do really cool ciruits and that's the fun part that's where I come in in uh in D wve so this uh image we see here it's a super conductor Loop a loop made of a super conductive material but it's broken in the middle now at higher temperatures this is just a broken Loop but when you we cool it down below uh the critical temperature of this material this becomes a very uh inter interesting object and I there's a schematic represent representation down here we have two elements or this ER squid will have two elements one is the inductor an inductor it's um it's a parameter that will appear anytime we have a a wire if we send current Through the Wire that will create some magnetic field and the energy store in that magnetic field we can can be represented by this element uh an inductor and the other element is the jstone junction it will be created by this break now we're going to take a look more into this but I like just to mention some funny uh I have this colleague he likes to say we do the best broken Loops in the world I like that example it's a because our CS they look like Arab squits and so we do broken Loops so this is the Joseph Junction this is the element that I showed before and as I said we have two superconductor materials that they come very close together but there's this isolating material between them now there's some very cool physics that I'm not going to get into it but at the when we are below or critical temperature there's going to be some current that is will be able to trespass this barrier this current we'll see it in this equation here will depend on the critical current of this Junction or IC and this is the maximum current that can exist through the junction and will be defined by the material and the geometry now the material and part of the geometry will be set by our fabrication stack like how we fabricate the junction and me as a designer I will choose the area of this Jun Junction and I can play with how much current can go through it the other parameter it will depend on is this phas this pH drop across the junction which is related to the wave function of each of these two uh parts of the loop but I'm going to share with you some book at the end and you can look through all the maths behind that if you are interested which I would recommend it's Prett interest it's nice and fun but what I want you to take from here is that a current will go through it depends on the pH drop and the uh IC critical current now let's go back to our eror squid or broken Loop let's imagine that oh per se it's just a loop with a with a junction but we can add a control we can bias it let's imagine we sent one wire all down our fridge and this wire goes very close to this Loop we send some current through this wire and that current is going to create a magnetic field that will be picked up by our Loop so we're going to call this magnetic field that it's picked up by our Loop or flux bias we are going to represent it like this in our schematic now we want to know this flux bias what kind of current it's going to create in this Loop we know that the current through the junction we gave it before it's dependent on the IC and a phase what uh this flux by is going to do it's going to be changing our phase so what Solutions we have to this schematic those are going to be the the Minima of the POS of the potentials um of the energy potential of our a of squid so we need to find the potential energy we can plot its landscape and the Minima of that landscape it's going to give us a phase number that we can plug into the into the our current uh equation and find what's the current around our Loop now we have our RF squid has two elements our uh inductor which is going to have this uh energy stored on it and you can see that it's going to depend on it's going to be a parabola shape if we look to it through the our face and our Junction which is going to have this energy here now if we sum both of the energies we're going to get uh one en potential energy landscape that I'm going to show some examples now but I want to point out that this energy landscape is going to depend on our two design parameters the IC of the junction and the L of the body so me as a designer I can change these parameters and design different behaving devices and it's also going to depend on the flux that we are putting into the loop so let's take a look at two potential landscapes in this example I chose an L an inductance of the body and an IC such that it has a mono stable potential there's only one solution to our uh AR of squid uh I think might be not very visible but there's a parabola in green here which is the energy of the stored in the inductor and there's a Sano here which is the energy stored in the junction so the sum of those two it give us this monostable potential in this other case I choose different design parameters and our landscape becomes or it has this multiple stable Solutions so two same circuit two different behaviors now let's look at a different kind of landscape in here we have what uh we called the double well so this landscape has two stable Solutions so let's go further and let's see what that means in our Arrow squid we take these two phe numbers we find from our stable Solutions we plug them into our Uh current equation and we will find that these two stable Solutions there's the same great or they are the same value current but in opposite directions now applying the right hand rule for those who prefer to talk about magnetic fields we're going to have a magnetic field pointing down for the left case and a magnetic field pointing up on the right case and some people might already be recognizing this those are our two classical states of our cubits so erf squid is kind of a simple the most simple version of our uh Cubit but we are not there quite yet we need to take a look at a different device this one is called the DC squid and this time it's loop with two Josephson Junctions not one and we also bias it through uh through some external using some external current and if we do that we will see that this device acts like one Joseph Zone Junction where the IC it we can change the IC we can tune the IC given this uh current uh this flux bias that we put into the loop so we take this this is squid and we will sorry yeah we will uh replace our Junction with this is squid we're going to name this now a compound Josephson Junction sorry for the change here so we have our compound Joseph Junction instead of the single Junction we had before and this is going to allow us to tune the IC of our era squid and these two uh external bias that we're going to use are going to allow us to um shape our landscape so we have one of these biases it's going to B go towards the body of our Arab squid and when we sweep it it's going to create this tilt this tilt is lowering the energy of one of our Solutions so it's creating a preferable State the other control is this flux V we put into the cjj this head of our Cubit and this is going to change our landscape from a monostable potential to a by stable potential this uh we call it raising the barrier and what it we are doing here is we are nailing the Cubit so let's go bit more into this what do we mean when we unal the Cubit so we start with the cuit in the with a single minimum potential it's a known ground low energy State and that happens with our two controls at zero now we raise the barrier we sweep the flux bias we put into the cjj of our Cubit and if there's no other external bias or cuid has a 50 50% chance of ending equal in each state well equal probability to ending in the zero or the one state but if we do put some flux bias into the body of the Cubit we are going to create a more probable State and now we're going to go back to the concept of the hamiltonian that we mentioned before and we can see that the A's and B's will be will evolve with our cjj we are going to anal using the flux bias that we put into the cjj of our cubid and the ages of our hamiltonian they will be specified by the flux that we put into the body we're going to be creating a more preferable State now we are missing the JS the JS we mentioned they represent the dependence of one cubit on the state of a different Cubit how do we Implement that this is an example that would do exactly what we want but it's not quite there yet and I will explain why so here we have a cubit which is coupled to another Cubit through a transformer for people not familiar with a Transformer this just means that if I put a current through here I'm going to get a current going out here which is uh value is going to depend on this parameter m so that's exactly what we want we want to be able to influence one cubit by putting by the state of a different one but we want to influence it in both ways right we want positive JS and negative JS I want to influence them to be in the same state or to the opposite State now here I use the What's called the not notation for do notation for the transformer for those who are not familiar we could just say that these two Transformers are the same but they have a negative M now again we want to implement this but we want to be able to switch between these two possibilities so we need to create a device that allow us to tune this m this coupling between these two cubits and that device it's another another eror squit looking device it's a very veral uh element but in this case in we are designing it so it has a mono stable poti and it's going to create an effective M an effective coupling between this Cubit I and Cubit J uh we're going to tune the M by sweeping the flux that we put into the cjj of the coupler so here we have a plot of an an example of how this device would behave for lower flux bias into the coupler we have a positive M so we have we would have certain U opposite or the cubits will tend to have opposite States and for higher biases of this uh coupler we would have a negative M and they would tend to be in the same state so this is how we Implement our JS or how we program our Js in our processor so again we're going to have if we have qit I and qit B uh cuid I and qid J sorry we're going to specify the hes of our problem by setting some flux biases into the bodies then we're going to specify the J between these two Cubit or the J between these two cubits by specifying the or by sending bias to the coupler the cgj of the coupler and then we're going to anal them by using the bias to the cjj of the cubits and if we swe this cjj bias if we anal them slowly enough the final value should uh be in the lowest energy of our system and we would find the the solution or we would find one of the Optimal Solutions of our problem now when we have two coupled cubits we're going to have four different states or Cubit each Cubit has two states and so we can see that if we add more cubits the number of states we can look at or that we can optimize is uh it grows exponentially with the number of cubits so more cubits bigger problems so let's make our processor bigger then here we have a layout that of what would be one of our previous processors you can see a series of squares this squares is what we call a unit cell and a unit cell has multiple cubits in this case eight cubits so in real life this is what I'm working on I'm throwing the layouts for like this would be an example of what I usually work with this in this layout our cubits are these extended Loops skinny long Loops that go from side to side of our unit cell now those uh cubits Are Not Alone they have multiple devices hanging on them maybe uh you can see that there's lot of boxes and other things going on in between them we use these devices for control calibration and readout and when there's a Crossing between each of these cubits we we are going to find there a coupler that's going to couple those two cubits that are crossing and we're also going to have some couplers that couple the Cubit to the next um unit cell Cubit so in this case this Cubit is going to couple to one 2 3 4 four five and six other cubits think that the more couples we have the more complex problems we're going to be able to solve those couplers is how we uh where the C couplers are going to be tied to JS right and the JS are kind of like the relationships or uh it allow us to build more complex uh or to solve more complex optimization problems now once we start growing and we start putting a lot of cubits together there's some design that poses some many design challenges and I would like to talk of some of them or at least the ones that we find the most or I encounter the most as a designer so we have two cubits we're going to need to put them kind of close together because we want to fit a lot of cubits in our processor now what I put a line down to bias one cubit it creates this magnetic field which won't go directly to where I want it will go everywhere and it might be picked up by an ed Cubit so imagine I wanted to bias this Cubit towards positive H and this Cubit towards a negative H but because of this cross talk there something might go wrong so we want to take care of that of these cross talks we have two different methods to do that first as a designer I can build walls between the cubits we told before that one property of superconductors is that they don't let magnetic flu penetrate at least not um there is a bit of a lie in what I said they do magnetic flux does penetrate a little bit but our walls are big if they are big enough they're not going to penetrate and we are going to isolate uh or keep it from from Cross talks another way is to compensate it in calibration and I'd like to use this chance to introduce what is calibration so here in thewave we have this big team of physicist who after we fabricate the Chip And we cool it down they're going to start talking to it they're going to start measuring what's happening there and they might measure some of these cross talks they're going to maybe start with this Cubit and they're going to say this Cubit is uh sensing some uh fil that it shouldn't sense there some cross to here then we are giving them as I mentioned before some programable devices so they can measure this cross talk and set program the device so it sends an opposite uh flux um bias into the body and does it compensates these cross talks another challenge that we find is due to process variations so when the chips are fabricated uh they are so small that there's some VAR small variations in the process that we they affect the parameters that we designed so the cubits might end up looking or having different design values than the ones we designed for and these designs values might depend on the location so if two cubits in our processor are far away from each other they might have different parameters so for example if the layer thickness changes from one side of the processor to the other we might find that our cubits they have different inductance body inductance that we told before so we add these devices called L tuners which our calibrators are going to measure our cubits they're going to see some of them have different inductance and they're going to program these devices to compensate for and homogenize the inductance of all of our cubits another example would be the the IC of our Junctions I mentioned before that as a designer I choose the area of our Junctions and that's going to set the IC parameter what happens is the area of the Junctions might change a little bit because of the U the fabrication process and so that might create what what would be kind of like a cross talk nonlinear cross talk into the body and that's really difficult for our calibrators to compensate so what we did uh is like I I don't know if you remember but two compound just one Junction acts like one Junction where you can tune the IC so we're going to replace those Junctions with two comp more compound comp compound J Joseph on Junctions now even though all these three ic's might be different because of process variations our calibrators can tune them to look like all ic's are the same so now that we went through what multiple cubits look like like to introduce you to our architectures so the one we were looking before it's what we call the chimer architecture here you can see the eight cubits are this long Loop uh Loops each Loop has a DOT where there's a coupler so there's one two three four five six couplers as I mentioned before and I as I said more couplers means we can uh have more complex problems also more cuits bigger problems so we are current advant such processor has 15 couplers per cubit we increase that and also increase the number of cubits and you might also notice that the cubits are kind of shifted so if this blue cubits are here and they finish here you can see the lower row of cubits is a bit stared it's shifted with respect to this row which is not happening in this topology and this allow us to implement different kind of graphs here uh there's a different visualization this is more towards when we are thinking about graphs and here the dots are cubits and the lines are the couplers this is the graph that we can Implement in Kimera topology and this is the graph we can Implement in Pegasus topology so again more cubits bigger problems more couplers more complex C problems and different uh yeah different drawing of the CU connections of between cubits different graphs now in uh so my team as I mentioned before it's called the processor development team and we are working in new generations of uh of processors we released the prop but we are working currently in what's going to be the seir topology this is going to have 5,000 plus cubits and 20 couplers per cubit so we're going to be able to solve even like more complex uh problems and also you can notice the layout it's a bit different too there's a different kind of a stagging or of location of the cubits yeah and very exciting of what new problems this uh new topology is going to allow us to solve so with this I'm going to finish my presentation and I'm going to open for questions but just before uh first thank you for everybody for listening and I'm going to add some additional resources here some books for who wants to go a little bit more deep into what the super conductivity is and how it uh works and also some link if you want to sign up for a free leap account and also if you want to participate on Friday's session with my colleague Sara I think you could already go there and get all ready I would really encourage you to join Sarah's Sarah's session thank you very much for listening to me fantastic thank you for your great talk Berta an Applause from everyone I see in the videos here online and also on Discord we will add Sarah uh Sarah James your colleague also to this Q&A so we can all uh you can both reply to the questions to briefly introduce you Sarah welcome uh Sarah jamus is a senior technal cont developer and instructor on the technical advising team at uh bewave and is responsible for creating and delivering training courses on Friday she will be back with uh talking more about the applications of the d-wave con and maner showing a demo and uh share more with you how you can do that yourself I will dive into the questions and uh first question is about these applications what what would uh what is quantum analing more suited for than the gate based quantum computers or is that the case is is it more suited can you share verta and Sarah yeah so I think there's a different applications definitely for gate model and anink so anink is more suited for as I said optimization problems as well as sampling problems and uh maybe s do you have some more in insights into that sure uh as Bera said basically we are focused on optimization problems uh the main like highlights that we focus on are M manufacturing such as Industries for example we look at a mod automobile industry and we have some life science we look at some chemistry like molecules and stuff behind the and we do some classifications for that we check some scheduling for example like employe scheduling nursing scheduling uh for example we do the tv ads scheduling and all of these you guys can see that are optimization problems that can be used on our Quantum and healing wonderful then back to the hardware an early question by Victor from Mexico um can you explain again more physically what happens if you CH if you apply changes to the hamiltonian in real life in the hardware and then samon from Malaysia asks how slow uh how slow is the evolution if you do the quantum and knealing in real time okay so um so that's what I was trying to uh show with the changes of the landscape I think that's really what is happening when we are kneeling we are changing the landscape so we are sending a current that into our processor that biases the cjj and body of the Cubit so uh yeah that's really what's happening in the hardware and sorry second question was how long does that process take okay so there's a different um uh the ranges say between the hundreds of the nanc and microsc and there is a question you showed you use superc conducting cubits um I see yes uh Victor from Mexico says that he also knows that iron trapped cubits could perform Quantum ining is that right could you in principle do Quantum ining with different Cubit modalities and is super conducting Cubit the most suitable one I think it is true don't uh maybe I'm not the best person for that as a this is a a question I would pass to my physicist colleagues but I do think that uh super conducting cuits are good in the way that we can scale them really easily and also there's already a lot of Industry that we can take advantage into manufacturing a lot of already we can take all the manufacturing information from the semiconductor world and what it's been learned before so was it's kind of easy way this to choose the super conducting cubits do you have some more insights Sarah maybe uh I don't have anything else to add this also shows how important it is that your team is composed of so many different people elect theorists um software Engineers a question when you showed the hamiltonians um Mohammed RZA from Iran says why the choice of poly X for the transverse hamiltonian and po Z for the pro one um he says I would imagine this works with any two sets of Poes as long as they don't commute okay so the idea is that at the beginning there's this um combination of uh all the or superposition of all the classical um classical uh states of our cubits and the X uh base if we have our cubits on our X base state it is a superposition of these two classical States so it is intended that it's on the X base we want to our initial hamiltonan needs to be or we want we prepare it to be in the lowest energy state of and it involves a superposition of all states so we want it to be there on the X and the final hamiltonian or program hamiltonian it needs to end up in classical states in the Z base so we can be able to read it out classically then back to Applications aad asks what is the biggest or most impactful problem that has been solved by d-wave so far and uh was it an implementation where a qu where your Quantum iner actually gave an Quantum a Quantum Advantage compared to classical computers think this one is for you Sarah yeah that's a good question uh we have a lot of uh examples we're going to talk more about that on Friday but you guys can see we have a lot of applications on our GitHub repository uh organization page which is like uh the gwave examples and uh some of the big ones are like portfolio optimizations we're going to do this example on Friday H some of them are employee scheduling for example some people are looking into seeing what's their uh favorite shift and we're going and you can see one of the repository will show you like how you can literally schedule People based on their preference shift um we have a lot more more um uh we have some paint chop optimization that we use uh that Volkswagen used as well and uh we they use the quantum and kneeling for that uh as you know like uh few of the problems can be solved on classical but it will take much longer to run Quantum and kneeling will like literally be much faster and sometimes we can get uh answers with millisecs when you use the quantum and kneeling um and again you can see all these applications and we're going to apply them on Friday it is really great and can you give some estimation how much faster is it some people ask yeah um Oh you mean between the classical and um right to be honest the classical sometimes take months so we take months to get results and when you use the quantum and leing you can get them in M seconds uh now we have two types you can send some something directly to our qpu or you can use our hybrid solver um when you use the hybrid solver you can have much more variable that you can Implement we're going to talk much more about that on Friday but uh by using these two techniques it will give you milliseconds and much faster than classical great and I just want to acknowledge more people for very good questions Laura from Spain also about the analing process and the time in it evangelia from Greece also about different types of cubits Mahia from Iran also and are multiple uh questions about the design um from maram from the US and also from cat Pakistan um um what software tool are you using to design the chip and is that open source can people use it so we are using cavens for the layouts that's a very uh typical software for the semiconductor industry too it's uh yeah it's not open source unfortunately and uh we do run some simulations with some open source uh because like the design is not just about drawing the layout right we once we have the the layout we need to see that it matches our schematics and then we need to run uh some uh spice simulations on those schematics to verify that the objects we design they perform in the way we intended and so that we there's open source W spice for example which can run uh it can run spice simulations on superconductor devices like just some Junctions great then uh a question from Muhammad why do you use a ring geometry I think that was when you showed earlier about the squids and around those slides okay so uh uh the ring geometry was as an example because it kind of gives you a very uh obvious idea of how the ring it's picking up the magnetic field so I'm sorry if that c caus uh some confusion we showed then how it's this a skinny Loop the cubid and we want it to be an extended element because we want it to cross different cubits in order to create a coupling between them and uh yeah the then the cjj the head of the cubid it also looks a little bit different in order to mitigate cross talks and also to like just what we are allowed to do with our uh multi-layer stack great and a question from Mandana from Canada close by maybe it's related to that uh Mandana says you mentioned more couplers uh cause more complexity um and what are the techniques uh yeah what are the challenges and what are the techniques to get rid of them to get rid of the sorry yeah Challen okay yeah related to couplers and the the complexity and the challenges they bring okay so for example as I said we intend to uh couple to cubits which are crossing together if we would do kind of like elant El elongated element as a coupler we would increase the body inductance of that ARF squid and it might not then it wouldn't have the monostable potential that we want it to have in order to behave as a coupler so maybe it would become this multi-stable state uh potential that we saw before and we could trap some flux in that so we wanted to keep it small so then we need to elongate the Cubit and that might bring us different problems so there's some yeah this an example of a design Challenge on the C wonderful then O'Brien from Kenya asks what are the limits of your Hardware in terms of Cubit coherence and perhaps also connectivity so we have a uh we are always working on improving our coherence uh coherence do I think it affects differently like Dan it does to gate model so maybe we are a bit more um permissive we can run problems even though our coherence time is shorter than I think it would not be possible in a gate model but we are working on increasing our uh coherence by Design or also by materials we are investigating materials a fun question from Victor have you ever used d-wave to optimize some internal challenges in the company or perhaps for the hardware for new topology so maybe it's just a fun fact but we use our our processor to do a Lottery Christmas uh giveaway but uh I don't recall using it in uh yeah not to develop other typologies how do you see the future of quantum analing versus gate based Quantum Computing I know that dawave is also developing a gate based quantum computer or maybe making it hybrid with Quantum analing I'm very curious what you can say about that yeah we are we are uh we have our gate model effort and uh we had already some designs like we are we even released the some white paper on how we could we are like our coherence on one kind of gate model Cubit uh I don't know if it will become a hybrid I think we are thinking them as different applications and it's a way that we can cover more customer uh like we can offer more services what's your input Sarah yeah and can you share a bit Sarah where would you see the different different applications from both sure um it's coming up so keep an eye on the news not maybe not not very like soon but uh as Berta said like we are trying to um build gate models and um I don't have a lot of input myself on that but um we are still building that and uh it's going I don't think it's going to be a we're trying it might be a mix of both but we are still in the process of building that and sah can you already give a sneak preview to Friday there is a question by o from Morocco how will we construct hamiltonians for optimization problems will we walk through a grid what will we how do we uh structure something like that can you give a brief preview sure on Friday we're going to go over two examples uh one of them is going the first one is going to be the antenna selection so we're going to learn how you can if you have an optimization problem how you can like um map it to a graph and then how we can basically access our qpu um and basically just run everything on the qpu and uh this is going to be directly this problem is going to be directly implemented into the qpu um and we're going to also do another problem which is going to be the portfolio optimization where we're going to use our hybrid uh solver and uh we're going to also go go over this demonstration now if you guys would like to literally follow every step with me I would highly suggest as Berta said like to sign up for a a free leap account that way you can have access to our uh cloud service and uh you can follow the steps that we will be doing together and uh if you would like more information you can check our repository on U d-wave examples and uh we're going to go over antenna selection and portfolio optimization to start reading before we start uh going over those on Friday excellent this uh the current banian program is mostly focused on Quantum but also on AI and Ben Tran from the US ask us a question about that uh Ben asks can we use quantum miners to solve or improve machine learning optimization algorithms do you have Insight on that yes um that's an experimental data uh like basically an experimental area we have been using few things like related to machine learning uh such as like the feature selection the classifications uh they are not directly part of the machine learning but they also work in adjacent uh way um as I said like uh some experimental we did our classification support vectors um machines and others in the machine learning space and uh don't forget we usually emphasize our um problem to be more optimization problems than for uh Quantum machine learnings better Sarah thank you very much for this exciting talk and for the comprehensive Q&A was really great uh lecture Vera and excited to have you back on Friday Sarah to uh to dive much more into the applications and algorithms we would love to keep in touch some people ask for your email address Bera it was visible on the last slide um which will then also be in the recording soon please feel free to contact me if somebody has some doubt and I can clarify them fantastic and thank you so much Marlo this has been great to give this talk hope it helps some people yeah and the very last thing I see some more questions here popping up what kind of people actually work at dawave and what kind of skills and backgrounds are needed so before we close it could you share a little bit more about that of course here in D wve there's a like I think a very wide amount uh variety of uh skills we have from like mechanical engineers electrical engineers to lot of physicist and mathematicians computer science and of course HR like we need from all kind of backgrounds I think there are some job postings open so of course go and look at that and there's always new ones coming up and uh I just want to say it's a great place to work I really love it uh I know you've probably noticed I don't know everything about Quantum Computing but I rely on my colleague and they are so welcome for me like to go and ask them questions and uh it's a great place to work and learn so I encourage you to check it out if there's some opening fantastic anything to add there Sarah yes I will add that you don't need to have a background of Industry to join thewave I'm coming from Pure Academia before here and uh literally I was so interested in all type of optimization problems I was kind of curious because for uh my PhD Theses every time I used to run my problem it used to take two months and a half to get the results and then I got too interested in Quantum and the quantum word and I was like maybe this will take millisecond but I was too late after I finished my PhD I started becoming more Curious and that's why I joined thewave so um keep in mind you don't have to have only industry background to join as long as you are interested you have some type of optimization problems backgr around um we welcome all type um of U jobs fantastic thank you again verta Sarah and looking forward to having you back on Friday Sarah then we enter the last 10 minutes of today's session I will walk you a little bit through what is coming up regarding exercises certificates you can obtain the flow of the program from now