 All right, so I guess we all have our coffee and tea awards. And yeah, so today, it's my good day having to welcome Ben Chau from Zürich, from Switzerland, from DCH to visit us. So I guess many of us already have really nice discussions and conversations and looking forward to the talk before that. Then we just say a few words. So Ben Chau studied physics in China, in the United States, but then right after that, I moved to the United States in 2011. So yeah, yeah. To do a PhD at the University of Illinois, to develop their experiments on quantum simulations with golden atoms, with gold fermions, and optical atoms, so to formulate the Fermi-Habbert model. And maybe one of the other, it was an observation or demonstration of many body localization in the system and many other things that we don't have the time to go for here now, but after that, around 2018, Ben Chau moved to MIT to start totally new experiments, do the Woodback states, for example, to do zoned linear optics and quantum linear optics. There's something that was quite a few years very much interesting in our center, but also to use the Woodback states to let us do this again, to set up applications in quantum simulation again, but also for quantum computing. That has been super successful, and it's now becoming a really big field for many other groups are following in the process worldwide. And so they were one of the first, actually, to establish the Woodback states. But actually, the success and the following success she now moved very recently to Switzerland, to the ETA, to the Paul Davos, too, as I was just done. Now to establish her own group and set up her own experimental program. And I look very much forward to the talk, and perhaps you also learn. Structuring and in vitration. And this is actually my first time to be in Denmark, and I really enjoy the discussion I had yesterday and also today. So it is my great pleasure to talk about, sorry, I'm just trying, yeah, okay, this way. Quantum science with readable atoms from one to many. So I started my tenure track assistant professor position at ETA from August 1st, last year, and I also have a joint position at PSI. So okay, let's start. So I think maybe all of us here know how quantum science and quantum technology is attracting the interest, not only from academic institute, but also from industry. So by harnessing the features of quantum mechanics, quantum device can outperform their classical version, like the application ranges from quantum sensing, quantum communication, then towards quantum simulation and the quantum computation. And I will say maybe as one of the ultimate goal, we would like to build up a general purpose quantum computer which can be used to solve some practical problems that cannot be handled with the most even powerful classical supercomputer. So to keep making progress along this direction, what I'll demand is to have large scale quantum systems, but also have the controllability at the single counter level. There are many promising experiment platforms that can be used to implement quantum technologies. I list some of them here. I will say each platform has its own advantages, but also face some challenges. So in my talk, what I will focus on is the experiment platforms based on neutral code atoms. And to be more specific, I will focus on the techniques based on neutral code atoms trapped in optic tracer arrays. So here is the outline of my talk. I think first for those of you who are not very familiar about how to use arrays of individually trapped atoms in optic tracers, I will give a brief introduction on this architecture and how we can use it for quantum computation and simulation. So in the seventh part of the talk, I will discuss a kind of alternative approach which was developed when I was a post-doc at MIT and Harvard. And in the last part of my talk, I will outline what I plan to do for my own group at ATH and the PSI. Okay, now let us start the story. And we know in principle, we can encode spins or qubits with any quantum objects which has more than two distinguishable quantum states. And for neutral code atoms, we can encode qubits and spins with two different electronic state of that atoms. And by applying some external electromagnetic fields, we can coherently manipulate the state of these electrons. Atoms have one advantage, which is they are naturally identical. So that makes them relatively easier to scale the system size up. And today, there has been well established techniques that can generate defect-free arrays of individually trapped atoms. So the basic procedure is as following by using some optic device, such as this special light modulator here. We can generate arrays of optic tweezers with any arbitrary geometry. And each optic tweezers is a tightly focused Gaussian beam and it can hold one and only one atom inside. So with this, you can assemble individual atoms together to get a ray of atoms with any configuration. And this configuration can be chosen depends on the problems of interest. And today, if you're, for example, Google atom arrays online, you can find many beautiful and interesting figures like people can even make after your tower in a 3D optic tweezers arrays. And to have this after your tower, this, of course, is done by a fragile group. So it's Antoine Browley's group near Paris. And you can even make funny cartoon movies by using this arrays of individually trapped atoms, like this Super Mario movie is met by Misha looking's group at Harvard. So all of these are at turning, right? But to make things more fun, you need to have interactions between atoms and notice that this distance between atoms is typically larger than one micrometer. So now the question is, how we can have strong interactions between these atoms if the distance between them is like larger than one micrometer? So the one of the approaches that we can have strong interactions between these atoms to make the physics, you know, beyond a single-part physics is to use the so-called Redberg atoms. Redberg atoms are highly excited atoms. So the wireless electron of that atom goes to an energy level with very high principal quantum number N. So it's really excited. So this wireless electron now will have a very large orbit radius and is very far away from the atomic nucleus. To give you a comparison, for ground state atom, the orbit radius of the wireless electron is typically at the order of 0.5 nanometer. But if you go to a Redberg state with the principal quantum number around maybe 1890, the orbit radius can reach 0.5 micrometer length here. So you can see there is a factor of thousand difference regarding how far away this electron is. And not because this wireless electron is so far away from the atomic nucleus, Redberg atoms can have really large electric dipole moment. And we know the interactions between atoms can be induced by their dipole-dipole interaction. For example, between two identical atoms, their interactions are one-wilder's interaction. For ground state atoms, one-wilder's interaction strength is so weak and is completely negligible if two atoms are micrometer away from each other. But the one-wilder's coefficient, C6, scares as n to 11, n is the principal quantum number. So you can already tell if n goes to 9,300, this factor is huge. So that's how we can achieve strong dipole-dipole interactions by exciting atoms towards their Redberg states. So now combining such strong dipole-dipole interactions with this array of atoms, we do realize a programmable counter-processor. There are two major ways to use this array of atoms as counter-processor. One is analog, one is digital. So quick summary about how to use this as analog counter-comp simulator is like in this scenario, what we used as our computation results is a many-body Hamiltonian realized by the arrays of atoms. For example, one can choose to encode the spin like spin down with ground state of that atom and spin up with a Redberg state. So in this case, you realize a quantum spin model which is a many-body Hamiltonian and you can use this to do some quantum simulation. The advantage of this approach is it is relatively easier to scale the system size up, but the downside is also kind of obvious as you can already tell, this is not a general-purpose quantum computer because you are restricted by what type of Hamiltonians you can realize. On the other side, for the digital approach, the computation is done by implementing a series of quantum gate operations like the following. And in this approach, people normally encode qubits or spins with two different ground state of that atoms because they are long-lived atomic state. And the Redberg state is only used as an intermediate state to facilitate some multi-qubit gate operations. This approach in principle is a general-purpose quantum computer, but the challenge at this moment is this approach is quite demanding or how well you can control your system. So this is my short tutorial kind of like how to use arrays of individually trapped atoms as a programmable quantum processor. And the past few years have witnessed a rapid progress on using it for study interesting physics and explore new applications. So examples include for quantum simulation, we can study quantum phase transitions like antiferromagnetic phase has been observed in 1D and 2D system and the quantum spin liquid has also been observed recently. And towards the direction of general purpose quantum computation, high fidelity to qubit gate operations have been demonstrated in multiple groups. And entanglement transport has also been realized. So here with this technique you can actually move atoms around in order to generate long-range multi-qubit entanglement. In addition, by combining some optimization algorithms these arrays of atoms can be used to do quantum optimization. So here we can actually use this platform to solve some classically hard optimization problems such as finding a maximum independent set on a graph. So in short, you know this as a programmable quantum processor arrays of individually trapped atoms have demonstrated its greatest scalability. The record of qubit number is about 320 now and also has great controllability over the positions of atoms. So that's really nice. However, you know just like any classical computer having a processing unit itself is not enough. You also need to have the input and output device. So what I mean here, well to be more specific we need to do state quantum state initialization at the beginning and also to detect the state at the end in order to know all the results, right? So for neutral atom-based platforms one thing kind of special is we actually need to construct this quantum computer at the beginning of each experimental cycle which means we need to use lasers to cool down these atoms from the background gas trap them and load them into optic tweezers, sorry. And this stage typically takes maybe 50 to 100 minutes at the time scale. After we're loading these atoms into optic tweezers some extra work is actually needed. This is because loading one and only one atom into an optic tweezers is a stochastic process. As you can tell here, some optic tweezers are actually initially empty. That is to say, in order to achieve the target defect-free configuration, you need to drag the atoms maybe outside the region of interest into the region of interest in order to form this target configuration. And as you can tell for two dimensional atom rate how to rearrange these atoms efficiently is actually not a trivial task. This rearrangement also typically takes about 50 millisecond. After we get this desired configuration we can use this array of atoms for doing some quantum processing. This can be done fairly fast. The gate is actually maybe less than point while microsecond and the total time maybe around 10 to 100 microsecond depends on what you want to do. So this is a faster processing. So now it comes to the readout stage. So here our goal is to detect the state of atoms. And the most widely used technique is based on single atom fluorescence image. The basic procedure is as following. Let's say we want to know which atom is spinning down which atom is spinning up. What we will do is we actually will first repair one type of atoms outside optic tweezers. Let's say we kick out the atoms as they are spinning upstate. So these optic tweezers become empty afterwards. Next, we can send some near resonant image light towards the remaining atoms and use a camera to collect the scattered photos. Now we know if we have bright spot on a camera the atom is in their spin downstate and if we have dark spot on a camera then it must be in the spin upstate. So that's how we read out the state of atoms. This image stage is kind of slow. That is because the absorption cross section of a single atom is kind of tiny. So in order to collect enough number of photons to have good signal to noise ratio, we need to wait some time so the detection usually takes few millisecond at least. And another big downside of this method is it is destructive. As you can already tell, one type of atoms are forever gone. Which means if we want to do a second next row on this control processor, we have to restart the experiment sequence from this very beginning construction stage. And now you can tell, okay, the time we are actually spending on the control part is even less than 0.01% of the total time of our experiment sequence, which is super inefficient. And in addition to that, because you have atom loss it kind of forbids you to do quantum error correction if you look from long term. So that's obviously two bottleneck of this current situation like the initialization and the readout, namely our speed. So we want to improve these two stages. So the motivation to improve these two stages motivated us to propose another different way to do things in which we are using a race of atomic examples for quantum simulation and computation. So here with a very similar, basically the same technique we can still generate a race of optic tracers with arbitrary configuration but now the difference is instead of having one and only one atom in each optic tracer, we have a small atomic cloud in each tracer. And as you can see from this cartoon, each tracer contains a few hundred atoms inside. And now for the encoding, we choose to encode our spins with different types of Redberg states. So here are the blue and red are different type of Redberg states. They have different angular momentum and principal quantum number and we can level them as spin up and spin down as this shows. So what are the advantages of using a race of atomic examples? Well, first of all, you actually have a faster generation of arrays. There is no need to do the arrangement. If catching one and only one atom into optic tracer is like using chopsticks to catch one ball which has a high failure rate, our approach is like using a spoon just grab a few hundred. And we even don't care the exact number of atoms we have in each optic tracer so there's no need to do the arrangement. And the second and the third advantage are we can have faster state initialization and a faster state readout. These two advantages, we first demonstrated with a single atomic ensemble experimentally which I will share in the following slide. So here is how we do the state initialization. We first know the few hundred grand state atoms into an optic tracer and now our goal here is to initialize this atomic ensemble towards this spin up state which has a blue type Redberg state inside. So this can be done by sending some night beams towards atomic cloud with the frequencies of these night beams match the energy separation between this grand state and this up state. So we can populate a Redberg extension inside. Now some of you may start to wonder what if you got more than one Redberg extensions in a cloud what if you have something like this which is even not in our defined computation basis, right? Well, the reason we can avoid this from happening is because we choose to work with a very small atomic cloud. So the radius of this atomic cloud is about three micrometer and because they are small the strong Redberg-Redberg interaction forbids us to get more than one extension inside due to this energy piloting. So this is known as the Redberg blockade effect and because of that we can guarantee we have one and only one Redberg extension per atomic cloud and the preparation time is about three microsecond and the fidelity is 93 for this work. That's how we do state initialization. Next question to ask is how we detect different Redberg state? So the goal here is we want to distinguish these two keys. The inverted detection scheme is based on a phenomenon known as electromagnetically induced transparency. So what we do is as following we're sending two coherent light beams towards the atomic cloud and these two coherent light beams couple the ground state of atoms towards this spin-down Redberg state through this intermediate state. Now if both light are on resonance with the corresponding transition the whole atomic cloud is actually transparent towards this detection beam so that's why it's known as electromagnetically induced transparency the EIT effect so it's transparent. However, now if you already have a spin-up state, Redberg state inside the cloud then the strong interaction between up and down will induce some energy shift so it will push this state up, for example. Now you can tell the EIT condition is no longer satisfied and because this EIT condition is no longer satisfied the whole atomic cloud become opaque towards this detection beam. So now you can see by monitoring the transmission rate of this detection beam we can know what type of Redberg extention we have inside the cloud and you can see this few hundred ground state atoms here amplify our optic signal which speeds up our detection speed. So here is a histogram plus the detected photon number on this detector for this two case so this case is opaque this case is transparent as you can see we can very well distinguish these two case so the fidelity of detection for this is 92% the detection time six microsecond is almost the effect of thousand faster compared to this conventional single-atom fluorescence image technique so that's a great news for us and we also do some analysis try to understand where the infidelity comes from and here is our analysis this plot shows the transmission rate of the detection beam as a function of the detection time for the case that you have a blue type Redberg in and without any Redberg in so in an ideal world both curves should be flat because there's no reason like the transmission changes but in reality they do change with the detection time especially for the case where you have a Redberg inside the slope has an obvious positive slope and we know this is because we started to lose this Redberg atom inside the cloud during detection the loss is due to actually the detection light so the light induces some Redberg state loss and the loss rate is about 0.035 per microsecond which contributes to this unwanted loop on this histogram plot and this actually hurts our detection fidelity because we cannot have longer enough detection time and the above slope slightly decrease this is because during detection there is some finite probability we create a Redberg impurity inside so that's why this slope slightly decrease and the creation rate of Redberg impurity is about 0.15 micro-select so that's the two major contribution to the infidelity so yeah that's our understanding on where the infidelities come from last but not least we also want to do some curate operation so here's the energy separation between the spin-up and the spin-down state is at like a few gigahertz frequency the energy difference so these are actually nice frequency to work with because there are so many electronic devices available and this plot shows the ruby oscillation so what we plot is the spin-up population as a function of the external microwave driving time and you see these nice ruby oscillations many many times and the time we need to flip the spin from up to down is about 19 nanoseconds and the coherence time measured through the Ramsey measurement is about 15 micro-seconds so as you can see the ratio between these two is not bad which means we can do many operations before this qubit decoheres into a classical bit so that's what we demonstrated so with this we have demonstrated that we can improve the state initialization and read out speed quite a lot and this paper was published two years ago on PIA so encouraged by this result we did actually a major apparatus upgrade afterwards in the hope to generate a race of atomic examples here are just some figures for fun what we do for our vacuum system and that is what apparatus looks like afterwards quite a complicated multi-layer because we want to mount a lot of 9-optic device but in any case that's the key optic components after this major upgrade we can now use a special night monitor combined with a high numeric aperture objective lens to generate large arrays of optic tweezers and trap atomic ensemble inside and in addition to this kind of hardware upgrade we also started to use some machine learning-based optimization algorithms after this upgrade we are using this open source Python package called Mloop to optimize our experiment loading and cooling so with this we actually improve the atomic number we can load into optic tweezers by a factor of three compared to human-based optimization and I will say another bonus of this you know in addition you have a factor of three improvement is this thing can actually be automatically done which means you can sleep at home and you just let your machine optimize it itself over light and next morning you get a happy machine so that's very attractive I guess especially for students and another unexpected gift actually given by this machine learning-based optimization is we actually observed a formation of both ice and condensate in optic tweezers with simple molasses cooling and the formation of BEC with simple molasses cooling for air-clad atoms has actually never been seen before experimentally and we somehow just kind of automatically get lot so we are writing a small paper on this surprise gift but I want to go into any details because it's kind of off the topic but if you are interested in this we can chat afterwards okay let's go back to our main story the arrays of atomic ensembles so with this we can generate arrays of atomic ensembles with any arbitrary configurations and now you can see here each branch part is a small atomic cloud has 200 atoms inside and the radius of this atomic cloud is about 1.5 micrometer I also want to emphasize all these images are taken directly after loading atoms into this optic tweezers arrays so there's no rearrangement involved at all so that's what we got that's the arrays of atomic ensembles and now for the detection what we will do is we will send a global detection beam to one's arrays of atoms and use a camera to do parallel readout and this is a very very preliminary result I got just maybe two weeks before I left MIT so what we do here is as following for the blue histogram correspond to the case without readable qubit inside the atomic cloud and the original one has a readable qubit inside and as you can see there's some difference regarding how many photons you can collect it on the camera if you have readberg or if you don't have readberg but the distinguishing ability is not as good as it was I guess it's better now but yeah that's what I got just before I left MIT yeah so that's a preliminary demonstration that we can do parallel fast readout of the readable qubit okay so that's kind of a quick summary of the second part of my talk where it significantly improves this state initialization stage and also the detection stage I also want to mention in addition to simply speeding up the detection another bonus we got is our detection is not demolished so atomic cloud is still here after detection which means if we want to start next round on this content processor we can directly start from here instead of from this stage so we can have much higher experimental repetition rate compared to previous approaches so yeah that's what we got and the next I guess we need to figure out how to improve the preparation and the detection fidelity because now we are going towards a larger scale system we also want to mitigate the cross talk between qubits in different atomic ensembles and I think this platform is very suitable for doing quantum simulation on quantum spin models so I think this concludes the second part of my talk and I will maybe pause here if there is any questions you could, you mean across ensembles right? Yes we could Yes that's definitely how it is You mean the size of the atoms that's basically determined by the temperature of the atoms and also the trapping frequencies so that determines the size of the atom cloud Yes it's a global beam That's a good question how you have individual controllability that's actually a general question that you can ask any people who work with a result of individually trapped atoms which I will try to explain what is my plan to resolve this individual and the flexibility, difficulty in my proposed architecture Yeah Maybe then I will also I think similar to what you just talked about with single atom I have this enormous asymmetry that you can use to monitor acting on so you can strongly interact with the data that's all you need to hang in there You know like here you form a few different states which are not so different that you work with a state-dependent interaction like a state-dependent much bigger way So we choose to initialize them actually to the P state Riederberg P-P are not that strongly interact if they are like few micrometer away you can kind of think they are non-interacting to some extent if they are that far away but self-blockage if they are like within one micrometer is still there but yeah It's like eyeballing Yeah, yeah, yeah, yeah, yeah, right, right So I think by encoding them with Riederberg states one advantage you do have you have more degree of freedom regarding how you want to them interact Yes Yeah Okay That depends on what is the Weblis Yeah, so because you basically as you can already tell here these arrays are generated by let's say this objective lens So it's basically diffraction-limited So it's diffraction-limited by you kind of naively think it's like determined by this Weblis divided by the numerical aperture of this lens Let's say this is typically less than 0.5 and your tracking Weblis is typically let's say 0.8 micrometer So the separation between these tweezers is the 1.5 micrometer So yeah, that's a typical lens scale that you can have Yeah, yeah, yeah, so that's I guess we also discussed that So that's why I think maybe Ethereum is better because you can start with some initially high state which the last excited states are more closer in energy So they have longer Weblis in transition That might be easier to fulfill your sample Weblis requirement In this is just the rubidium So it's typically the first transition is 780 nanometer Yeah, nothing too special about these atomic spaces Oh, but no, between red and black it's like a gigahertz So it's a centimeter Yeah, yeah, right, yes, right Yeah, yeah, yeah, yeah, right Yeah, yeah, yeah, right Yeah, yeah, right Oh, yes, so we actually don't want to be easy So the funny thing is like 20 years ago people tried so hard to get busy and the father will even don't want to be easy because the state is too high So we avoid how easy in the tweezers Yeah, yeah, yeah So that's why we are kind of careful about what is the beam of waste of the optic tweezers If they are too tight, it's not good Yeah, yeah, okay, cool If you know further question I will move on to what I plan to do for my own group Yeah, okay So, yes, so let's take a look at the wish list for quantum computer, right So I think these items will be there So I will say Artitecture based on neutral items trapped in optic tweezers can get a great score regarding scalability and also coherence time If we consider the gate fidelity at this moment maybe the highest one is like maybe 98.5% gate fidelity which is not as high as like ion based platforms for example but we are catching up But I do think there is a lot of space to be improved regarding the programmability So what I mean by programmability is we actually need to figure out how to fulfill what I listed here regarding like local qubit operations on a subset of atoms which I will bring it up and also how to do non-demolish and selective qubit detection and also reuse qubit for high experimental reputation rate So that's what we need and I think with all these we may be able to develop a path towards implementing quantum error correction So for me, my goal for my own group is I want to develop new schemes that can fulfill these needs here and use my platform to study new physics and also new applications To reach these goals what I plan to do is to build up I call it hybrid quantum system So what I mean here is I plan to combine array of individual atoms and array of atomic ensembles and here is a schematic diagram showing what I propose to do as you can see here there are two types of arrays The bottom one is made with arrays of individually trapped atoms and I plan to encode my spin or qubits with this bottom layer and the top layer is an array of atomic ensembles and I have some ideas about how to use these atomic ensembles to do some auxiliary operations on the below individual qubit and I can have controllable interactions between these two layers the method I will discuss maybe in a few minutes So regarding the choice of atomic species for the bottom layer I plan to use the terminated atoms Oh, yeah, okay, sorry Yeah, so I won't bother you with atomic physics too much but the tag-home message is the terbium atom as the air-clad earth-like atoms it has two valence electrons instead of one and because it has two valence electrons it actually has many nice features as listed here which makes it become a very promising building block for this atom-based architecture and for the atomic cloud I will simply use rubidium atoms Rubidium atoms may be the simplest atoms to work with in our cold atom community so rubidium are our best friend The reason to choose different atomic species for this dual-type array is because I want to avoid unwanted crosstalk between these two layers and also with this I can actually have more degree of freedoms regarding how well I can control internal interactions and internal interactions To have a controllable internal interactions what I plan to do is to use the so-called foster resonance which allows me to tune the interactions between these two layers with external electric field To see the working principle let's take a concrete example about a specific choice on the Redberg state For example, now I choose to use this one of the wireless electrons of the Eterbian goes to the 62 S state 62 is the principal quantum number S is the angular momentum and my rubidium has this Redberg state at 71 S Without external electric field there is almost no interactions between this Eterbian Redberg and rubidium Redberg But what will happen if I started to apply some external DC electric field? Well one thing for sure is we know the total energy of this pair of Redberg states will change as a function of external electric field which is the stark effect so it changes as this plot shows the blue curve is the total energy of this Eterbian and rubidium and it kind of goes down as we are turning on the external electric field But since we have become more fun if we also plot the energy of some nearby Redberg pairs For example, now the closest one regarding energy is the Eterbian 62 P and rubidium 70 P so as shown by this original curve and as you can see at some finite electric field the blue and original cross each other like the energy state become degenerate here But because there are dipole-dipole interactions between this choice and this original choice and this blue choice it will actually open a gap here to have anti-crossing and this induces interactions between Eterbian and rubidium so that's the forced resonance And what this plot shows is the interaction between these two Redberg states when they are about 5 micrometre away from each other and the interaction strings near the crossing spot is actually comparable to the Redberg-Rubidium interactions between two identical rubidium atoms But the advantage we have now such interactions can be rapidly switched on and off by simply changing the external electric field So it's really flexible regarding how you want them to interact or not So that's my plan to have controllable interactions between these two layers And I believe because now I have controllable interactions between these two layers I can do many, I can demand many new skills to fulfill what I need to do here Let's start with this selective non-demolished detection first because the working principle is basically the same as what have been demonstrated at MIT So now what we will do is as following in order to know what type of Eterbium atom we have here we will monitor the optical response of this atomic cloud So the detection beam actually couples atomic transitions in rubidium atom cloud and if you encode a qubits with ground state of Eterbium atoms you can excite one of it towards the Redberg state before you do detection Now because you have interactions between this Eterbium Redberg and rubidium Redberg the optical response of the atomic cloud condition on what type of Eterbium you have, right? So that's how we achieve mega detection But one major advantage of this proposed scheme compared to what we have done at MIT is you can not hear the qubits of interest especially separated from your detector and all the detection beam is actually unresolute with Eterbium ion the Eterbium kind of stays in dark during the detection stage and because of this we are actually expecting a much high fidelity is maybe about 99% fidelity based on our estimation and I think selective detection is also possible because all these scattered photos are not unresolute with Eterbium so it won't decoher as the Eterbium are not interested in So this is how we achieve this selective and non-demolish rapid detection and because of this detection scheme we kind of naturally can reuse qubit for high repetition rate and also combine these two, we at least fulfill the minimal requirement for doing quantum error correction So now the only thing remaining here is how to do reconfigurable local qubit operation or subset of atoms So what I mean here can be maybe kind of illustrated by one simple example So what I want to do here, let's say I have a way I only want to do some qubit operations on this free for example at T equals what? and then to do something else on this free you know, not a global operation This, you know, first maybe you think it's quite a simple go, right? The easiest way you can do this you can simply send in some like qubit operation beams with their pattern, match the pattern of qubits of interest then these three qubit evolve and the other qubits remain unchanged where although this looks quite simple on our PowerPoint slide it's quite hard to realize that experimentally this is because from the experimental side to generate this arbitrary pattern of optic beams with strict regulation on their power and frequencies is almost impossible for currently available commercial optic device So this won't work, that's just a two challenge for available device So as an alternative solution what we can do is we can have local light shift by applying some far-off resonant individual addressing beam towards this qubit So this green light actually coupled one of the qubit state towards some other intermediate state far away from resonant and through axis dark shift the qubit with individual addressing beam will have actually different energy separation compared to the qubits without this individual addressing beam but this individual addressing beam is far away from resonance so you kind of relax your strict regulation on the frequency control that opens up the possibility of using current available optic modulators However, the downside of this approach is okay sorry I forgot to mention now you can send some global qubit operation beam towards a race and some you know respond some don't because they are off resonant but the downside of this approach is because now you make some short-lived intermediate state into your qubit state you for sure reduce your qubit lifetime and you also limit your qubit operation speed because operation speed cannot be faster larger compared to the induced local energy shift and the high power on this addressing beam is always preferred causing some additional technique complexities so that's the downside of this approach so to overcome these challenges here what I plan to do is differently I want to use my atomic ensemble to assist some individual qubit control what I could do here is I can selectively create readable gas stations in this atomic cloud which can be done actually technically not hard because for this purpose I even don't care the generation of readable gas station inside the cloud is a coherent process or not so that's quite easy to do and now as you can hear afterwards I can use this readable atom inside the cloud to index the below interbium atom so now even with global qubit operation beams these three qubit will respond differently to most global qubit operation beams we have many some different ideas about how to implement one qubit operation and two qubit operation I guess we need to try to see how well they perform I just want to quickly mention one I'm kind of excited to try this is a single qubit operation scheme and it is inspired by this theoretical proposal actually many years ago for different purpose but we try to use some spirit in mentioning this paper so what we do here is we are sending three light beams towards the interbium atoms which actually fulfill an EIT condition without bothering you with details the tag-hole message is if you keep the control light on all the time and you ramp up and ramp down this omega p if the EIT condition is satisfied this qubit will actually not involve it stay unchanged however if now you have a readable extension in the nearby atomic cloud let's say you have this case then the interaction between this rubidium readable and the interbium readable will destroy the EIT condition now in this case your system kind of go back to a textbook example of two-fold-time Raman translation and your qubit will start to rotate with the rotational angle equals to the area below this omega p so that's what we want to try for individual qubit control okay so that's we completed this and we hope, our hope is by combining one single atom and many atomic, many atoms nearby to get the best of both and with this we hope we can make this shooting on cat happy and yeah so that's what I hope to improve on the technique side and I do think this dual-type binary structure could be a very promising new architecture to explore new physics and also new applications so for the scientific, okay I have five minutes for the scientific part what I plan to start initially is to investigate quantum information dynamics so this topic you can think it collects actually many different fields from fundamental physics to application you know examples including quantum civilization long-equilibrium dynamics even quantum gravity and how to do topological encoding and because generally speaking I started these problems with classical computers it's hard because the here ball of space is simply too big to be simulated efficiently so there are still many open questions and experimented platforms you know like trapped ion superconduct superconducting circuits and atom arrays are really good platforms to investigate these unknown open problems what I am particularly interested in is to use my platforms to study the situation where the unitary quantum evolution is interspersed with some measurement so what I mean here is like you have a system you can do some subsystem measurement during the quantum evolution and you study how the subsystem measurement changed the evolution afterwards this topic has recently attracted a lot of interest from theories for example on the fundamental physics side people discovered a universal quantum phase transition which is induced by this measurement so this quantum phase transition is really sharp the entanglement entropy changes from a volume law to a scaling law so if you do like a high rate measurement the entanglement entropy will scale according to an area law and on the applied side people do think this thing has a deep connection with some threshold theorem for achieving fault-tolerant quantum computation because you can view this measurement as some types of errors so there must be some connection here and the hope is by starting this we may get better understanding on how to dynamically protect our information by doing dynamic encoding but this topic is really hard as far as I know there is no concrete proposal on this is because you're not only encoding itself is not enough you also need to decode this efficiently so that's hard but what I do know is by combining measurement and some classical communication it is possible to implement some measurement-based protocols which can do quantum information processing more efficiently so that's what I plan to try in my group after I get my apparatus running and I also want to point out although there are a lot of interest from theoretical physics doing this study experiment is still quite a challenge regarding this quantum phase transition so far there are only two experiment papers from trapped aisles and also superconducting circuit there is no experiment done with neutral code atoms mostly due to the difficulties of doing long-demolish subsystem measurement and also reuse of qubits because my platform can mitigate these two challenges I'm really excited to start this topic okay so this concludes my talk and so the work has done at MIT and Harvard is like a joint project between Vladimir Tich from MIT and Misha Lutin from Harvard and those previous PhD and wasting master's students and those current PhD students and for what I discussed today for my ETH group I have a master's student at ETH and he helped me doing some next simulation on the foster resonance and I also have some PhD and post-doc students in my group now and I do have a post-doc position open so if you are interested or you know who may be interested in just contact me yeah thank you for your attention and I'm happy to answer more questions