Kai's Headshot Figure showing three-level Λ-system used in some quantum memory implementations Figure showing an SEM image of a nanofabricated SiN waveguide Kai Shinbrough

PhD Student

University of Illinois–Urbana-Champaign


"[A]ll physical theories, their mathematical expressions apart, ought to lend themselves to so simple a description that even a child could understand them." —Albert Einstein, conversation with Louis de Broglie.
[R. W. Clark, Einstein: His Life and Times (1972).]

I'm interested in boiling down complex scientific ideas and making analogies with everyday experiences in order to make those ideas accessible to everyone. In my experience, science isn't actually all that complicated; scientists just tend to obfuscate simple concepts with jargon, and focus on technical details that really aren't all that important. In an attempt to clarify the extremely small subset of science that I have been involved with, below you'll find summaries of my scientific papers written in "layman's terms."

K. Shinbrough, T. Loveridge, B. D. Hunt, S. Park, K. Oolman, T. O. Reboli, J. G. Eden, V. O. Lorenz, High-Efficiency, high-speed, and low-noise photonic quantum memory, arXiv:2309.00969 (2023). Submitted to Nature.
Alright, this is the big one from my PhD. We worked for 3 years building an experiment that relies on a high-power laser [O(1 mJ), 1 kHz, 100 fs pulses, for the experts], a high-temperature oven (800-900 °C — more than 1500 °F), and a high-pressure vapor cell (1300 mbar — OK, actually, that's pretty close to 1 atmosphere, which is not that crazy by human standards, but is very high for most atomic experiments) to create a quantum memory in neutral barium vapor. Barium is a solid metal at room temperature, and only melts at 727 °C, hence the need for the high temperatures to create an appreciable vapor. Once all that hard work was done, we were able to tickle our barium atoms in just the right way with our laser and a high-pressure of argon buffer gas that the atoms could accept and store an ultrafast single photon. And when we wanted the photon back out, we were able to tickle the atoms in just the same way, and retrieve the photon with roughly 30% end-to-end efficiency. Now, 30%, might not sound like much, but for the field, and for photons this short in duration, it's a big deal. We have a plot in the first figure of the paper comparing our result to the rest of the literature on these atomic-ensemble-type quantum memories, and this work represents a significant breakthrough in efficiency and bandwidth for this type of memory. The memory also comes with a number of side-benefits, like ultra-low noise and telecommunications wavelength compatibility, and we were able to perform a new characterization method for this type of memory that allowed us to measure the "shape" of the photon coming out of the memory.

K. Shinbrough, D. R. Pearson Jr., B. Fang, E. A. Goldschmidt, and V. O. Lorenz, Broadband Quantum Memory in Atomic Ensembles, Adv. At. Mol. Opt. Phys. 72, 297 (2023). Invited Chapter. DOI: 10.1016/bs.aamop.2023.04.001.
Quantum memories are devices that catch photons—particles of light—and release them on demand. As you can imagine, there are several important metrics for quantum memory. Efficiency is one good example; it's not so useful to have a quantum memory that only stores and releases 1 photon for every 100 photons that come into the memory. That's an example of a low-efficiency memory; instead, we want high-efficiency memories that store and release every photon that comes in. What we focus on in this work is a tradeoff that exists between memory efficiency and memory bandwidth. Memory bandwidth describes how 'fast' the memory is, i.e., how many photons you can read in and out of your memory per second. It turns out that quantum memories tend to be either highly efficient and slow, or fast but low-efficiency. In addition to comprehensively reviewing the literature on this subject, we provided some physical intuition for the efficiency-bandwidth tradeoff, and described the possible paths forward to overcome it.

K. Shinbrough and V. O. Lorenz, Variance-based sensitivity analysis of Λ-type quantum memory, Phys. Rev. A 107, 033703 (2022). DOI: 10.1103/PhysRevA.107.033703.
Quantum memories based on gases of atoms have been around for a while, and there are lots of theoretical and experimental papers on the subject. One thing that hasn't been investigated much, however, is the 'sensitivity' of these quantum memories. In the real world, quantum memories are subject to all sorts of experimental fluctuations and drift, and a memory that is more robust, or less sensitive, to these sources of noise is more useful than a memory that is more sensitive. In this paper, we investigated the sensitivity of a large class of quantum memories, and we showed which memory 'protocols' are most robust, and in which scenarios.

K. Shinbrough, B. D. Hunt, and V. O. Lorenz, Optimization of Broadband Λ-type Quantum Memory Using Gaussian Pulses,
Phys. Rev. A 103, 062418 (2021). DOI: 10.1103/PhysRevA.103.062418.

Photons are particles of light. Photons can store and transmit information, and they can even be used as the fundamental building block in a 'quantum computer' (albeit there may be better building blocks). One big problem with photons, though, is that they move at the speed of light—the speed limit of the universe—and that makes them hard to 'catch' and hard to use in practical devices for that reason. (This is not the only reason photons are hard to use, but it's a big one.) Quantum memories are devices that 'catch' photons, and release them on demand. In this paper we showed that one of the most common types of a quantum memory, which relies on clouds of atoms, is compatible with "Gaussian pulses." This just means we can use pulses of light that have a smooth, regular shape in time, and the quantum memory will work pretty well. [What does it mean for a pulse of light to have a 'shape' in time? Imagine turning a lightbulb on and off as fast as you can—we can talk about the pulse of light you just created as having a duration (however long the lightbulb was on), and as a function of time it has two sharp edges where you turned it on and then back off. What if you had a dimmer knob instead of a light switch? Then you could smoothly turn the light on and off, creating a pulse with a smoother 'shape.' Same idea here.] It turns out that this smooth Gaussian shape is easy to make in the lab, and it's a lot easier to make than some of the strange shapes that have been proposed in other papers. To be fair, those strange shapes in other papers do a little better than our smooth pulses—the quantum memory is a little better at 'catching' the photon—but we showed that the smooth pulses do almost as well for much less work, and we showed an interesting connection between different quantum memory 'protocols' that is revealed by making this switch to Gaussian pulses.

K. Shinbrough, Y. Teng, B. Fang, V. O. Lorenz, and O. Cohen, Photon-matter quantum correlations in spontaneous Raman scattering,
Phys. Rev. A 101, 013415 (2020). DOI: 10.1103/PhysRevA.101.013415.

It turns out photons (particles of light) are only really useful, in most circumstances, if they exist in "pure" states. What does it mean for a photon to be in a "pure" state? In this case we are considering photons that are emitted from materials in a particular process called Raman scattering, and we say a photon is "pure" if, just by looking at the photon, you can't tell anything about the material. Well, ok, that's great, but how would you ever learn anything about the material just by looking at the photon in the first place (i.e., why aren't all photons pure)? Well, in the case of Raman scattering, when you create a photon, you also create a phonon, which is a little vibration in the material. The photon and the phonon are only created in pairs; you can't get one without the other. This means that if I have a photon with a little less energy than normal, that missing energy must have gone to the phonon, and if I could look into the material I would find a phonon with a little more energy than normal. Ok, so if I measure the energy of my photon, I can learn about something going on in the material, i.e., how much energy the phonon has, and this means my photon isn't pure, and that's bad. How do I get pure photons from this process then? We showed that if you inject some uncertainty into the energy of your photon on purpose—you set things up so that you always get photons out that have a large spread of energies, no matter what is going on with the phonon—you can wash out any information about the material that would otherwise be encoded in the photon energy. So that's pretty cool! Inject some uncertainty, and the photon you get out actually becomes more pure, at least according to this definition of purity. But it turns out there's a problem with this scheme. When I inject this uncertainty into the photon energy, I also do something else. When I make the spread of photon energies larger, I also unavoidably make the photon shorter in duration due to a little something called the Fourier transform. Ok, so the photon you get out is shorter in duration... who cares? Well, it turns out that if my photon is really short in time—shorter than the time it takes for the photon to propagate through the material—another process comes into play that gives me information about the material, just by looking at the photon. This process is called "dispersion." It turns out that when my photon is propagating along through the material, it's not actually moving at the speed of light. It's moving just slightly slower. And since my photon is really short, that means that I can tell how long the photon propagated in the material by measuring when it arrives at my detector, which is placed downstream. If the photon was created at the very beginning of the material, it had to propagate through the whole thing before reaching my detector, and it got slowed down significantly and arrived 'late.' If the photon was created at the very end of the material, it hardly propagated through any of the material at all before hitting the air and moving on to my detector. So if my photon is too short, I get some information about where in the material its partner phonon is, and the photon is no longer pure. These two processes compete, and there is a 'sweet spot' in the middle where my photon is long enough that I don't know where along the material the phonon was created, and short enough that there is enough uncertainty in its energy that I can't tell the energy of the phonon. We showed that this sweet spot exists, how to calculate where it is, and we did some experiments to confirm.

Y. Zhang, R. Spiniolas, K. Shinbrough, B. Fang, O. Cohen, and V. O. Lorenz, Dual-pump approach to photon-pair generation: demonstration of enhanced characterization and engineering capabilities, Opt. Express 27, 19050 (2019). DOI: 10.1364/OE.27.019050.
This experimental work was mainly performed by my fellow PhD student, Yujie Zhang, and the theoretical work was mainly performed by Bin Fang and Offir Cohen. I will try not to butcher their ideas!

Above I claimed that photons are only really useful if they exist in "pure" states. Here we looked at a different process for creating photons—something called Spontaneous Four Wave Mixing (SFWM)—where you send two photons into a material and you get two different photons out. We normally say that the two input photons were 'annihilated' and the output photons were 'created.' The usual way people make photons using SFWM relies on sending in two photons that are the same color (the same frequency), and getting out two photons that are different colors. The output photon colors are both different from one another, and different from the input photon color. It's known that if you're careful you can get 'pretty good' photons out of this process; that is, photons that are 'pretty pure.' What we showed is that if you use two input photons of different colors, and you choose those colors carefully, you can get photons out that are totally pure, not just 'pretty pure.' What is going on here relies on the fact that different color photons propagate in the material at different speeds. I said before that photons are impure if you can tell something about the material they came from, just by looking at the photon. In the usual case, since the photons propagate at different speeds, it turns out you can actually tell how long the material is, just by looking at the photons! If the output photons were created at the beginning of the material, and one propagates faster than another, they arrive at a detector placed downstream from the material at different times. But if the photons were created at the end of the material, they don't propagate along the material very far, and the two photons don't pick up any delay between each other. If you know the speeds of the photons inside the material, you can measure the largest time delay between any two photons created at the same time, and this tells you the length of the material! Kind of a roundabout way of measuring length, I know, but you've got to admit that's pretty cool. The important part, however, is that this ability to measure the length of the material makes the photons impure. The idea in this paper was to use two input photons of different color so one photon is 'fast' and the other is 'slow.' If you send the slow photon in first, then the fast photon, the fast one will propagate along and overtake the slow photon while inside the material, and exit ahead of the slow photon. Since you can only get this annihilation and creation process when the two input photons overlap, in effect this choice of different colored input photons slowly turns the interaction 'on' and 'off' while the fast photon catches up to the slow photon and overtakes it. Since the interaction is 'off' at the beginning and at the end of the material, you can't tell just by measuring the output photons how long the material was! Because of this, the output photons are now pure. Yujie and I (and others) did an experiment to confirm this, and we did some pretty nifty noise analysis.

S. A. FitzGerald, K. Shinbrough, K. Rigdon, J. L. C. Rowsell, M. T. Kapelewski, S. H. Pang, K. V. Lawler, and P. M. Forster, Temperature-programmed desorption for isotope separation in nanoporous materials, J. Phys. Chem. C 122, 1995 (2018). DOI: 10.1021/acs.jpcc.7b11048.
Everyone knows about hydrogen gas (H2), but not many people know about its heavier sibling, deuterium gas (D2). Each hydrogen atom (H) has only 1 proton in its nucleus, and 1 electron orbiting it, but each deuterium atom (D) has 1 proton and an extra 1 neutron in the nucleus. It turns out that this makes deuterium very useful—its used in 'heavy water' (D2O) to cool nuclear power plants, it's used for chemical labeling in fancy chemistry research, it shows up in certain drugs that metabolize slower than the same drug does using hydrogen atoms, etc.—but D2 only shows up naturally in 0.02% of 'hydrogen' gas that's harvested or manufactured. If you want to use just the D2, you have to separate it out from the H2, and this is a really hard problem because H2 and D2 are so similar. Industrial methods that are used today are slow and expensive and inefficient. An alternative approach for this separation has been proposed that relies on 'nanoporous materials.' The idea relies on the fact that D2 bonds 'adsorbs' onto these materials slightly better than H2 does. What is 'adsorption'? Just think of it as a really weak attraction of D2 to the material, where D2 just sticks to the surface. Since this is a surface interaction, these 'nanoporous' materials—little sponge-like materials with holes (pores) that are nanometers in diameter—are a good candidate for use in this separation technique, because these materials have humongous surface areas for very little volume. This adsorption process only happens when the material and the D2 are really cold, around 4 K. So the idea is to take some of this 'dirty' hydrogen gas with some deuterium in it, expose it to these nanoporous materials, and cool the system down slowly to 4 K. The D2 and H2 should both adsorb, but the D2 should be more 'sticky,' which we call having a higher 'binding energy.' So if you get a bunch of H2 and D2 adsorbed into these materials and slowly heat the system, the H2 should be released first and voilà! You can remove the desorbed H2 and you should be left with purified D2 attached to the material, which you can then release into D2 gas by heating the system further. In order for this process to work, however, you need a large difference in the 'binding energy' of H2 and D2 at these surface sites. If the difference in binding energies of H2 and D2 is always some fixed percentage (say 10%) of the average binding energy for both, you would think that the larger your average binding energy, the better your separation will be. 'Better' separations have larger 'selectivity.' What we showed in this paper was the unfortunate result that this isn't the whole picture. If the difference in binding energies between H2 and D2 is always a fixed percentage of the average binding energy, at first you get a big advantage when you increase the average binding energy (the selectivity rises sharply), but after a while the selectivity stops increasing and it doesn't help to have a larger binding energy past a certain point. This was a result that came out of some computer modeling that I did of our system, and we did many experiments with different materials that had different binding energies to confirm the results of the model were correct.

S. A. FitzGerald, C. T. Eckdahl, C. S. McDonald, J. N. Nelson, K. Shinbrough, H. W. H. Lai, and J. L. C. Rowsell, Orientational ortho−H2 pair interactions in the microporous framework MOF-5, Phys. Rev. B 92, 134304 (2015). DOI: 10.1103/PhysRevB.92.134304.
Ah, my very first paper. This one seems, on the surface, very complicated. Lots of complicated formulae, numerical calculations, theory, and experiment. But in the end, like all the rest, there is a simple physical picture that explains everything we did. The motivation for this work has to do with Metal-Organic-Frameworks (MOFs), which are 'tinker-toys' for chemists. They are arrangements of 'ligands,' which are just straight-ish molecular links that connect 'metal groups,' which are sticky clusters of atoms much smaller than the ligands. If a MOF were shaped like a cube, the metal groups would be at each of the corners, and the ligands would be the lines connecting the corners. MOFs are great because chemists can mix and match different ligands and metal groups fairly easily, and you can make a really wide range of materials with different properties. One key application of MOFs is hydrogen storage. Hydrogen gas is a great fuel, and it's clean. When you burn hydrogen, in a car, say, all you're doing is reacting two parts of hydrogen (H) with one part of oxygen (O), and what comes out of the tailpipe is their molecular combination: water (H2O)! None of this messy CO2 and other junk you get from burning gasoline. So hydrogen is great, but there's a problem. It tends to explode when exposed to the air. This means that if your hydrogen storage tank in your car gets a pinhole leak (as it may in an accident, for example), it may explode. That's not great. What if we could store the hydrogen inside one of these MOF cages, so that the whole MOF-hydrogen system is intert—it doesn't explode—but you can heat up the system slightly to release the hydrogen from the cages and use it as fuel? That would solve this big problem with hydrogen fuel, and that's one of the key applications of MOFs. But MOFs are complicated, and the interaction between hydrogen and MOFs is much too complicated to model completely on computer. We know that hydrogen sticks to MOFs, but all the details are hard to work out. If we could figure out exactly what is going on inside these MOFs, we may be able to design bigger and better MOFs that are more suitable for use in, e.g., cars. That's the motivation for this work. We took a very well-known MOF, "MOF-5," and we loaded hydrogen into it. We then shined infrared light onto the hydrogen to see how its absorption of light changes when its inside the MOF. What we found is that in the case of MOF-5, hydrogen first sticks to an active site in the MOF-5 metal group, but as you keep loading in more and more hydrogen, the hydrogen at one site starts to interact with its nearest neighbor at another site. It wasn't known that hydrogen did this in the first place, and it further wasn't known how exactly this interaction of hydrogen with its neighbors works. We considered a number of different candidate explanations for this interaction, and we showed that, this time, the interaction in MOF-5 relies on the 'quadrupole moment' of hydrogen. Two hydrogen atoms can attract each other through this quadrupole moment, and we showed that if you take away the quadrupole moment by switching to a different form of hydrogen, the nearest-neighbor interaction goes away. Is MOF-5 a good candidate for real-world applications of hydrogen storage? No, not really. But with our newfound knowledge of quadrupole interactions, maybe we can design a better MOF in the future. More important to me, however, is the fact that we learned something about what is going on down there, and we now have an incrementally better understanding of how our amazing universe works. And to me, that's what science is all about.

Contact:  Kai Shinbrough Social Media:  LinkedIn Footer image of images from across the website
Office 347, Loomis Laboratory Research Gate
1110 W. Green St., Urbana, IL 61801 Google Scholar