WEBVTT 00:01.960 --> 00:08.679 And thank you for the organizers for this opportunity to discuss our work. Um, in this 00:08.680 --> 00:15.120 talk, uh, although I've enjoyed hearing a lot about diamonds and small scale magnetic resonance, 00:15.520 --> 00:21.360 this morning we're going to change topic and go to a different sensor and different length scales 00:21.959 --> 00:27.160 and talk maybe more about the themes of miniaturized, the size of the overall NMR 00:27.160 --> 00:32.280 instrument rather than the sample that we're going to study. And so that involves these types of 00:32.280 --> 00:38.240 sensors which are involving atomic spins, uh, atomic 00:38.720 --> 00:45.640 vapors of alkali metals like rubidium, cesium, etc.. If you want to know more about 00:45.640 --> 00:52.440 that, I can recommend that you consult this, uh, recently published review article, which was 00:52.440 --> 00:59.120 written by several of the people in the room Dmitri, John, James and myself and others, uh, of how 00:59.150 --> 01:06.029 this area, how this intersection of vapor based magnetic resonance sensing and nuclear 01:06.029 --> 01:12.790 spin sensing, uh, has developed. Um, and had this had this long, 01:13.269 --> 01:19.109 let's say, intersection over over many years, maybe approaching two decades now. 01:21.230 --> 01:25.469 Since I think that's unfamiliar to you, I'll give a kind of tutorial introduction to it. And then I 01:25.470 --> 01:31.910 want to explain more about, um, how this could go on in the future towards smaller sized 01:31.910 --> 01:38.429 systems. So I'll talk about, uh, small size hardware 01:38.430 --> 01:44.230 and samples, um, to generally increase the overall 01:44.230 --> 01:51.150 accessibility of these type of magnetic resonance experiments. And I'll give you an idea of some of 01:51.150 --> 01:56.030 the applications as well. And we work in this kind of feedback loop where when we develop some 01:56.030 --> 02:02.499 hardware, it leads to ideas for new experiments and keep. Keeps going. So I think that, uh, this is, 02:03.139 --> 02:09.779 um, a nice environment to discuss that in with many people here. Okay, so, 02:10.259 --> 02:15.499 uh, since I think this is unfamiliar to you, I'll show you what what an atomic magnetometer looks 02:15.500 --> 02:22.179 like. Um, this is the key sensor part. So this is, uh, a little box which 02:22.179 --> 02:28.179 contains an alkali metal. This is rubidium hanging out here, and there is some equilibrium between 02:28.179 --> 02:35.179 atoms in the vapor and the metal. It's oxygen free. Otherwise, the metal would simply oxidize. Uh, and 02:35.179 --> 02:41.020 then detecting magnetic fields involves nothing more than passing light through. Uh, so long as 02:41.020 --> 02:47.619 that light is resonant with the atomic transition and is is circularly polarized. So before going 02:47.619 --> 02:54.179 into detail of that, uh, we just see some examples here. So we could, we could either detect the 02:54.179 --> 03:01.089 magnetic field through the precession of the the electron valence spin in the atoms. 03:01.369 --> 03:07.929 Or we could detect just based on the transmission. Um, the transmission depends on magnetic field 03:07.929 --> 03:13.049 through a saturation effect, which you could see through the diagram. Kind of like this So the key 03:13.049 --> 03:19.889 thing is that if you have a you treat the electron, uh, just as a, as a single electron 03:19.889 --> 03:26.770 particle, you forget about the, the nuclear spin of the alkali metal. Uh, and you imagine 03:26.770 --> 03:32.809 there being a population here, which could be interconverted by a magnetic field. So this type 03:32.809 --> 03:39.569 of device, uh, as it's drawn here, involves a continuous pumping on the, the optical 03:39.569 --> 03:46.329 transition, uh, which transfers population over to this state here, but that it could be brought back 03:46.330 --> 03:52.329 again by the magnetic field. So if you're at exactly zero magnetic field, guess what? This will 03:52.330 --> 03:57.760 reach a saturation where all the population is here. And it leads to a maximum transmission of 03:57.760 --> 04:04.039 the light through the vapor. Whereas if you're at some magnetic field which is not zero, then this 04:04.080 --> 04:09.520 acts to reequilibrate these states. And then the vapor can always absorb more light. And you're 04:09.520 --> 04:16.359 more on this part of the curve. Um, okay. And then this is quite nice because it's it's very simple. 04:16.880 --> 04:22.799 It involves relatively few components. You just have your, your vapor and, um, 04:23.840 --> 04:30.719 light source photodetector and some coils. So an important thing to note is that this is 04:30.720 --> 04:36.199 an area of magnetic field sensing that's already gone through a MEMS revolution. We've gone through, 04:36.400 --> 04:43.000 uh, cells like this which have some flame seal. Uh, here, they're made on a kind 04:43.000 --> 04:49.480 of, uh, glassblowing manifold to something which can be made in bulk and this is 04:49.959 --> 04:56.909 a MEMS structure. So it has a silicone body. Um, this sorry is in 04:56.910 --> 05:03.589 centimeters. This ruler. Okay. Uh, how are these fabricated? They're done on a wafer scale, and it 05:03.589 --> 05:10.549 involves etching cavities in silicon. And then you plop on some some windows, and before you 05:10.590 --> 05:16.109 seal bond the whole thing shut, you will insert the alkali metal in a in an in an inert 05:16.109 --> 05:22.869 atmosphere. Uh, you bond all of that shut, and then you can cut this up with a dicing saw, and you can 05:22.869 --> 05:28.950 make many hundreds of these cells at a time. Uh, this there are there are various methods of 05:28.950 --> 05:35.910 introducing the alkali metal. I guess I'll just know that one of them involves, um, uh, is 05:35.910 --> 05:42.429 a little bit more friendly. That involves an azide salt of the alkali metal. So you have, uh, 05:42.670 --> 05:49.350 say rubidium azide or cesium azide, and you'll put a little droplet of that into, into 05:49.429 --> 05:56.379 the cavity and then you bond the the cell shut, and then you activate a 05:56.380 --> 06:01.419 thermal decomposition of the azide with UV light. And that generates the pure metal. And it also 06:01.420 --> 06:06.019 generates nitrogen gas which is a buffer buffer and quenching gas which is important for the 06:06.060 --> 06:12.659 these these reaching a high sensitivity. Um, okay. 06:12.700 --> 06:19.539 And this is uh, not not a new thing. So many, many groups and institutes have been developing these 06:19.540 --> 06:24.579 types of devices for different applications. I want to point out this one here. We worked with 06:24.579 --> 06:31.219 the Swiss um, Micro Fabrication Institute, uh, to develop this two 06:31.260 --> 06:36.540 chamber cell here. The reason why it has two chambers is that one is used for the azide 06:36.540 --> 06:43.459 filling. kind of a dirty backyard for for inserting rubidium into the cell. And then 06:43.459 --> 06:50.329 there's a clean chamber here, which is where we, uh, pass light through for magnetic field sensing, and 06:50.329 --> 06:55.409 there are some tiny micro channels that will connect these two. So even if there's, for example, 06:55.409 --> 07:01.249 incomplete decomposition of the azide, then it's matter. It's going to be left in in the backyard 07:01.249 --> 07:08.088 and it will not, for example, block up the windows. Okay. So this is all leading to something like 07:08.089 --> 07:14.890 lowering the cost, allowing us to integrate these into devices. Or maybe, maybe the fact that planar 07:14.929 --> 07:20.969 planar manufacturing is is more easily done by robots. Um, and that that helps the general effort 07:20.969 --> 07:27.570 of, uh, making things smaller. I want to highlight two innovations that that our group has made in 07:27.570 --> 07:34.409 this area. So one of them is to add extra functionality to these vapor cells. Um, here we 07:34.409 --> 07:40.929 had, uh, not just these type of cavity cells, um, with two windows, but we actually put 07:40.969 --> 07:47.409 platinum metal functionalization onto the windows. So you had heating wires, uh, 07:47.440 --> 07:52.720 Resistive heating wires and thermocouples that you could use to measure the temperature and 07:52.720 --> 07:58.719 control temperature gradients within the cell. So you could have a temperature gradient, which takes 07:58.720 --> 08:05.359 all of the rubidium access and condenses it in the backyard. So again the sensing chamber is 08:05.359 --> 08:11.639 clean, just like here for example. And this has been used 08:11.640 --> 08:17.838 in an atomic magnetometer. It reaches the kind of sensitivity that you would 08:18.520 --> 08:25.159 you would get with traditional glass blown vapor cells or without additional functionality. So 08:26.320 --> 08:31.239 there are some risks in doing this, this type of functionalization. What you don't want to do is 08:31.279 --> 08:38.080 add functionalization that also adds noise to the sensor because that's counter active okay. 08:38.119 --> 08:44.959 And another one concerns the coils. I mentioned that there is a zero field 08:44.960 --> 08:50.709 resonance phenomenon, which can be used to detect the magnetic fields. And that's one which is 08:50.750 --> 08:56.309 actually the most commonly used in the commercial magnetometers because it involves the fewest 08:56.310 --> 09:02.709 components. We developed a set of three, uh, three axis, um, 09:02.710 --> 09:09.149 PC, um, field coils that are just on two planes of a PCB. So you make a 09:09.189 --> 09:16.069 PCB printed circuit board, which has got, uh, traces on here, um, which will generate 09:16.069 --> 09:22.789 shim fields in, in X, Y, and z directions so that one can get this sensor, 09:22.909 --> 09:28.549 uh, operating at, at the zero field point, which is where it where it needs to be to detect the 09:28.549 --> 09:33.868 magnetic field. Uh, there are some constraints in this sort of design. Like you need to leave a hole 09:33.869 --> 09:40.630 for the light beam to pass through. Um, you also want to try and make this, um, set of 09:40.630 --> 09:47.299 coils as homogeneous as possible. Why? Because basically the T2 star of the the atom, the 09:47.340 --> 09:54.019 atomic spins is what determines the sensitivity or is what can limit the sensitivity. Um, 09:54.020 --> 09:59.779 so you could quantify that, for example through the linewidth of a free induction decay signal. 10:01.140 --> 10:08.059 And in this case there was a that design that was a design consideration where you want to minimize 10:08.060 --> 10:14.620 the, uh, inhomogeneity over the directions that matter. And you don't. And you can therefore 10:14.620 --> 10:21.619 afford to, uh, root traces, uh, over, over less homogeneous path for the direction that, that, uh, that 10:21.659 --> 10:28.419 doesn't really matter, which is along the beam. Okay. I want to say then 10:28.460 --> 10:33.659 that while these sensors are quite cool and they're very sensitive, um, they're not really 10:33.659 --> 10:40.019 optimized for, for NMR, um, they're optimized for a completely different application, which is, which 10:40.020 --> 10:46.849 is measuring the magnetic fields from, organs like the brain and the hearts, um, 10:47.210 --> 10:54.009 from outside the body So these are designed to be worn on the body as a, as a 10:54.010 --> 11:01.009 replacement technology for all type of squid sensors, which are used in, in MEG So in MEG 11:01.050 --> 11:07.129 you sit in this kind of while your head goes in the warm side of this cryostat. And then there's 11:07.169 --> 11:13.649 there are squid sensors. Um, the idea is to instead wear the sensors directly, 11:14.369 --> 11:20.489 which is possible because they're not cryogenic. Um, and then you can place them anywhere on the 11:20.489 --> 11:26.049 head that you like. Or people have different size heads, like kids and adults have very different 11:26.049 --> 11:32.649 size heads. Um, and this also accommodates moving the head rather than keeping your 11:32.650 --> 11:39.610 head fixed in a, in a in a certain position. So if you can imagine this type of, 11:40.210 --> 11:46.929 Um, brain imaging system in a hospital. Say, um, it gives you you can 11:46.969 --> 11:53.449 kind of roughly calculate how many sensors you might need to get started. Uh. Um, let's say you 11:53.450 --> 12:00.250 want 100 sensors, coverage over of the scalp, and then let's say you you're going for 100 of these 12:00.250 --> 12:07.130 imaging systems that's already, already going to be 10,000 vapor cells. And then, uh, 12:07.169 --> 12:12.769 you might need to have another factor of ten for yields. Um, so you're getting on for something like 12:12.809 --> 12:19.529 100,000 vapor cells. So then it's not only vapor cells, but it's individual 12:19.530 --> 12:23.968 magnetometer modules which are placed over the surface of the head. And there are a bunch of 12:23.969 --> 12:30.649 design considerations which I won't go into all of them, but many of them are similar to, um, 12:30.689 --> 12:36.289 what you might need if you were designing this type of magnetometer for NMR, like the fact that 12:36.289 --> 12:42.399 you're detecting a local field. So you want you want the atoms to be as close to the edge of the 12:42.400 --> 12:49.359 package as possible, rather than sitting somewhere in the middle. Um, and that's 12:49.359 --> 12:56.199 that's the same design criterion in The MEG because you want the sensor to be as close to the 12:56.199 --> 13:03.159 the surface of the skin as possible. Um, I'm going to get back to NMR in a 13:03.159 --> 13:10.080 second, but I want to say then that. So there are many, um, commercial prototypes of 13:10.080 --> 13:16.599 these sensors. You can buy them. They're quite highly optimized. Um, however, I think as you could 13:16.599 --> 13:23.478 probably see from the pictures here, that they don't look that, uh, able to be assembled by robots 13:23.479 --> 13:30.399 at the moment. Um, so we think that the innovations we've made could, if you design a magnetometer as 13:30.400 --> 13:36.599 a stack, which is completely robotically assembled, this would this would help in this area. Um, 13:36.599 --> 13:43.469 Also the the components that I showed you, uh, are also able to be 13:43.469 --> 13:49.989 made very, very cheaply. So the vapor cell and the field coils. Okay. Field coils is just a PCB. So 13:49.989 --> 13:56.949 that's extremely cheap. Uh, the vapor cells, you could imagine that being, um, I don't want to put a 13:56.950 --> 14:01.869 price label on it, but I think it would be about two orders of magnitude cheaper than per unit 14:01.869 --> 14:08.869 than you can currently get these vapor cells for if you make them at scale and therefore, uh, I 14:08.869 --> 14:15.389 don't think that, um, that buying these types of sensors and then trying to make an 14:15.390 --> 14:22.029 NMR instrument based off of this type of detection technology, um, 14:22.030 --> 14:28.589 is that smart of a move because, uh, basically this part of the sensor here in the middle is 14:28.589 --> 14:35.589 extremely cheap. Um, uh, it's like an insignificant cost compared to the rest of 14:35.620 --> 14:42.099 it. So the expensive part of these devices really is the laser at the moment. Um, the 14:42.100 --> 14:48.699 laser is about 90% of the cost of the device, so it's going to require certainly more than 14:48.739 --> 14:55.179 NMR, uh, to be interested. It's going to involve, I think, a lot more than MEG to be interested, to 14:55.220 --> 15:02.059 make these, these lasers on such a scale that, uh, that they become lower in 15:02.060 --> 15:08.699 price. These are certainly not made in the same type of volume as DVD lasers, which cost just a 15:08.700 --> 15:15.579 few euros each. So, um, we're talking orders of magnitude more than that. So I think that's a 15:15.580 --> 15:22.140 limitation. But if these types of magnetometers are going to be used in in small scale 15:22.140 --> 15:29.099 sensing, including NMR, then well that's a consideration. Okay. So I want to come 15:29.099 --> 15:36.048 back to NMR now. Uh, finally and um explore a little bit of a different direction, which is 15:36.049 --> 15:42.049 now less talking about the the sensor, but more about the type of system that we study with it. 15:43.250 --> 15:50.210 And so as we've kind of already heard, um, um, these types of 15:50.210 --> 15:55.209 sensors can be used in different ways. I usually like to show it like this, so that the already 15:55.249 --> 16:00.089 what's been done with alkali metal vapors and nuclear spins can be summarized with this picture. 16:00.530 --> 16:06.769 Uh, just the alkali metals by themselves as magnetometers or frequency standards. 16:07.329 --> 16:14.289 Um, many of you will know that you can use, uh, xenon or helium, uh, within the vapor 16:14.290 --> 16:20.489 cell to sense rotations and make gyroscopes. And also, if you have a way of 16:20.490 --> 16:26.849 introducing the the noble gas and taking it out of the cell without consuming the rubidium in the 16:26.849 --> 16:32.880 process, you can you can hyper polarize the gas and that's used in lung MRI and so on. And then 16:32.919 --> 16:39.839 the, the one which was mentioned in the previous talk, which is detecting the magnetic 16:39.840 --> 16:46.319 field from nuclear spins outside this box of atoms. So this is the this 16:46.919 --> 16:53.719 ZULF NMR detection technique. So it, it's limited by the ability to match the 16:53.719 --> 17:00.679 frequency of this system, uh, to the sensitive range of this system. So 17:00.679 --> 17:05.799 if you go to high in field, uh, the fact that these have different gyro magnetic ratios means 17:05.800 --> 17:12.598 that they're not going to crosstalk with one another. This was first done, um, 17:12.640 --> 17:19.079 around 2005. This example is from the Romalis lab. So they used a potassium 17:19.079 --> 17:25.559 magnetometer and they used, uh, water protons. And then again, thanks to the 17:25.560 --> 17:32.029 miniaturization which has been provided by, uh, MEG type wearable 17:32.030 --> 17:38.989 sensors, then this. This has got to the point where you can literally buy one of these. Get some 17:38.989 --> 17:45.469 sticky tape and you can sticky tape this to a solenoid, okay. Like this. And you can 17:45.469 --> 17:52.309 put some sample inside like water or, or some hetero nuclear system. And you can use that to 17:52.350 --> 17:58.749 detect either Larmor precision precession or some kind of J Coupling oscillation of the 17:58.750 --> 18:05.670 magnetization in the sample. I also want to mention that in this 18:05.670 --> 18:12.188 area there have been some other efforts. So there was a microfluidic, um, kind of 18:12.229 --> 18:19.229 flow cell. This is a remote detection, um NMR. It was also used for remote detection MRI with 18:19.230 --> 18:25.790 remote encoding. So you had a micro fabricated, um, cesium cell next to some channel 18:26.469 --> 18:33.179 and then a micro fabricated, noble gas polarizer. These were both 18:33.219 --> 18:38.819 at least ten years old. These works. And I think it's really a shame not to have seen more 18:39.499 --> 18:43.620 development in these areas. I think that they're really before their time. So if we were thinking 18:43.620 --> 18:49.979 about, um, low or low field NMR in a, in a chip scale 18:49.980 --> 18:55.019 setting, kind of along the lines that have been discussed from, from this network, I think that 18:55.020 --> 19:01.620 these ideas are well worth revisiting. Okay. What is the future? I want to talk about a few of the 19:01.620 --> 19:08.299 experiments that, um, that we've done in this area. So a few years ago we used, uh, 19:08.419 --> 19:15.139 magnetometers. So here's the rubidium vapor cell down here to measure fast field cycling, um, 19:16.220 --> 19:23.060 relaxation data sets. So from from proton. So if you have a sample which sits inside a 19:23.060 --> 19:27.890 polarizing coil which you can energize and then switch off and Then you can measure the 19:27.890 --> 19:33.849 relaxation rate as a function of the magnetic field, say between or nearly all the way down to 19:33.890 --> 19:40.609 DC and then up to say megahertz frequencies. And that was done on these 19:40.930 --> 19:47.569 kind of milliliter sized vials of liquid. So protons in water. And so we thought 19:47.610 --> 19:54.368 what would happen if we didn't use this type of sample. But we used something which was more 19:54.369 --> 20:00.849 interesting, more representative of a chip scale system. So this was a fiberglass 3D 20:00.850 --> 20:07.729 printed set of cavities. Um, we put that in the same instruments. Um, 20:07.729 --> 20:14.449 and the way to spatially resolve these cavities was simply to apply a field gradient 20:14.569 --> 20:20.889 along along the axis. So that resolved each of these little cavities into the peaks that you see 20:20.890 --> 20:27.800 here. Um, so it's it's still, say, very early, this set of experiments, but it's 20:27.800 --> 20:34.199 illustrating that small volumes are tolerable in in low magnetic 20:34.199 --> 20:40.919 field, uh, NMR with these types of sensors. Um, we could do something with contrast imaging or, 20:41.359 --> 20:47.079 uh, solvent relaxation, or we could add additional functionality like flow in these, in these types 20:47.079 --> 20:54.079 of devices, which you think is very interesting. Uh, we extended that to imaging. So 20:54.120 --> 20:59.359 in addition to the gradient along the field axis, then there was then a phase encoding along one of 20:59.359 --> 21:06.319 the other perpendicular axes. Um, so with these types of structures we could 21:06.319 --> 21:12.879 resolve water in the different cavities. Again it's not it's not sub micrometer scale magnetic 21:12.880 --> 21:19.399 resonance. Um, but it's magnetic resonance on samples which are very generous in terms of 21:19.400 --> 21:25.629 materials tolerance. And you could imagine putting all kinds of tubes or fittings, etc. 21:25.670 --> 21:32.670 etc. to make interesting devices. And we're also able to also able 21:32.670 --> 21:39.589 to show some kind of synergy here with the atomic sensors, because we moved the position of 21:39.589 --> 21:46.509 the sample with respect to the vapor cell. It's what these different images show. And so when the 21:46.590 --> 21:51.429 when the vapor cell is over a particular region of the sample, you're more sensitive to the NMR 21:51.469 --> 21:57.709 signal from that region. So here you see when the vapor cell is here then then there's you can kind 21:57.709 --> 22:02.789 of see the signal. And when it's down here and then you see this part of the sample, it's 22:02.789 --> 22:07.709 providing us some feedback on what the magnetometer is able to see. So we would be able 22:07.710 --> 22:14.269 to tune the size of the sample size of the vapor cell to the sample that's being looked at. 22:16.310 --> 22:22.899 Um, I don't think this in-situ prepolarization method is really the best one. It's, um, 22:23.339 --> 22:29.579 it's kind of slow. It involves lots of accumulations. So, uh, we, like many others, have been 22:29.579 --> 22:35.579 using shuttling system to move samples quickly between a high field prepolarization field and 22:35.580 --> 22:42.459 then the low field for detection. Um, back some time ago with Dimitri Sakellariou we 22:42.459 --> 22:49.420 made a permanent magnet, uh, which was just made of rectangular magnet blocks. And this reached a 22:49.420 --> 22:56.379 field of two Tesla. Um, for for this, uh, for, for this shuttling type of experiment that wasn't 22:56.380 --> 23:03.179 big enough. So we decided to increase the size of that magnet by a factor of two. Um, so these magnet 23:03.180 --> 23:09.858 blocks were no joke. They were one one inch by one inch by four inches. Um, but it was still it was 23:09.859 --> 23:15.299 still possible to put this array together by hand. And we think that we could, uh, maybe gain an extra 23:15.300 --> 23:20.259 factor of ten roughly in volume by by buying some bigger magnet blocks, another factor of two in 23:20.329 --> 23:25.290 size and making making some, uh, pre polarizing magnet. 23:27.489 --> 23:33.569 Okay. Um, I want to switch slightly again. Talk about high polarization 23:33.569 --> 23:39.889 hyperpolarization. Everything is better with hyperpolarization. Um, we use the fact that, 23:40.290 --> 23:45.728 uh, you can use a magnetometer to detect the DC magnetic moment of the sample. So without using 23:45.729 --> 23:52.649 any pulses. And the analogy here is just making a movie of magnetization, um, 23:52.770 --> 23:59.609 without with just looking at the spins and measuring, they're measuring their fields So if 23:59.609 --> 24:04.930 you have a vapor cell next to the sample and then there's some magnetization dynamics, then you 24:04.930 --> 24:11.489 could just look at that and the sensor will will see the signal. Um so we explored that 24:11.489 --> 24:18.489 with with para hydrogen. Um I think I won't go into detail on this in 24:18.489 --> 24:23.198 the interest of time, but there are some interesting magnetization dynamics that one can 24:23.199 --> 24:29.799 look at here and perhaps control or use feedback from hyperpolarization processes 24:29.800 --> 24:35.800 in real time. We also did this with SABRE. This is where 24:36.800 --> 24:43.039 instead of completely passively measuring the sample, we did something like toggling with pi 24:43.040 --> 24:50.039 pulses between plus and minus magnetization. Uh, and if you as long as you just 24:50.040 --> 24:56.759 toggle plus and minus plus and minus and back again, you can get a square wave signal. And then 24:56.760 --> 25:02.279 that lets you see the signal of some magnetization accumulation without without 25:02.279 --> 25:08.479 actually destroying it. And we did that also with dissolution DNP. So there are many examples there. 25:08.520 --> 25:13.078 We think that there are there are opportunities here for quality control of hyper polarized 25:13.079 --> 25:19.469 samples without without or instead of real time monitoring with these types of 25:19.469 --> 25:26.389 techniques. So there's clearly a lot of interesting applications of NMR at 25:26.389 --> 25:32.670 low field and. Applications and measurements with, with 25:32.670 --> 25:39.429 magnetometers. Um, it also should be in principle very simple. There are many, 25:39.430 --> 25:44.869 many, many motivations for that. You don't have many of the problems that you have in in high 25:44.910 --> 25:51.909 fields NMR spectroscopy. But a major block to that at the moment is the availability of 25:51.909 --> 25:58.749 instrumentation. So, uh, to put all of the operators together that you might 25:58.750 --> 26:05.510 need to build an ultra low field system, you might be kind of better of buying a high field 26:05.510 --> 26:11.749 system and doing some high field NMR with that, uh, maybe second hand. It kind of works out similar 26:11.749 --> 26:17.099 price. But as I mentioned then that the commercial op amps, we expect to see those come down in price, 26:17.420 --> 26:23.300 or expect to see those being significantly cheaper. If you don't buy something which is 26:23.300 --> 26:30.219 optimized for a different type of measurement. As for the console. So we developed that. We developed 26:30.219 --> 26:36.859 a console for uh sub megahertz NMR using Arduino. This is based 26:36.860 --> 26:42.619 on uh off the shelf component. So you can buy most of these from the electronics store. You can just 26:42.620 --> 26:48.060 put them together in the, in the, in the lab with soldering and everything fits on a credit card 26:48.099 --> 26:54.579 sized board. So you can just carry it around in your pocket. Okay. And we also shrunk the size of 26:54.579 --> 27:01.098 the magnetic shield. Um, the main cost of the magnetic shield is the, the material itself that 27:01.099 --> 27:07.139 it's made from. So if we, we half the size of the shield, uh, it becomes roughly a factor of ten 27:07.139 --> 27:13.059 cheaper. Uh, and then this one is about coffee mug size, so I like I really like the, the, the allusion 27:13.060 --> 27:20.049 to that in the, in the beginning of the meeting. Um, and then this is a simple enough system that we 27:20.050 --> 27:25.209 use it in teaching labs for, for students to learn about magnetic resonance and do pulse programing. 27:25.250 --> 27:32.089 So pulse programing is microsecond time resolution. Um, and you can also do simple 27:32.089 --> 27:36.609 experiments like you can program simple experiments like CPMG to measure a relaxation 27:36.609 --> 27:41.488 times and some, uh, hetero nuclear type experiments as well. 27:43.969 --> 27:50.769 Uh, okay. And we gave one of these to Ajoy's lab, and, um, this was around a couple of years ago when 27:50.769 --> 27:56.849 I went to EMC. I took one of these, and within one afternoon, we just completely replaced all of the 27:56.849 --> 28:02.930 electronics on, uh, ZULF type spectrometer that they were building. Um, 28:03.849 --> 28:09.889 got it all working. And a year later, then they produced some very nice results on multi-channel 28:09.889 --> 28:16.359 detection. Um, although still using these commercial sensors. But, uh, this is this is a very, 28:16.359 --> 28:23.079 very nice work. And, uh, um, they've come a long way with it. Okay. Uh, so 28:23.119 --> 28:30.080 I want, um, to kind of close this talk by, um, taking away 28:30.080 --> 28:36.919 that we can, we can use the fact that, um, these atomic, we leverage the fact that 28:36.920 --> 28:42.800 atomic vapor cells have been developed to a high technology because of other areas,like, like MEG. 28:42.840 --> 28:48.839 Um, and we could use those in NMR, but we could also use those in a, in any kind of other 28:48.839 --> 28:55.240 area of magnetic field sensing. Um, we also explored the possibility of just putting a vapor 28:55.240 --> 29:02.159 cell on a PCB. And so you can carry that around and put that in, put that in a physics experiment. Uh, 29:02.320 --> 29:09.279 this was a laser lock. So you simply have some hot vapor phase atoms here. It contains all 29:09.279 --> 29:16.269 of the heating type, uh, circuitary that you would need, uh, on the window of the vapor 29:16.269 --> 29:21.708 cell. it works just by finding the the absorption minimum of the vapor. And then you just 29:21.709 --> 29:28.509 lock to the you lock to that frequency. So it's as simple as inserting a lens into the beam, which is 29:28.510 --> 29:35.229 not not the case with things that you can buy these days. And then, um, I want to finish on one, um, 29:35.589 --> 29:41.948 unexpected thing that that came up in the course of these experiments, that ultra low field, 29:42.829 --> 29:49.669 uh, we so we discovered an interesting way of getting high resolution spectra. So, uh, 29:49.709 --> 29:56.510 you in very, very low magnetic fields where you have, uh, a crossover between strong coupling and 29:56.510 --> 30:02.389 weak coupling involving, uh, Zeeman Hamiltonian and J coupling Hamiltonian. You have what are called 30:02.389 --> 30:08.469 anti crossing fields. Um, so, uh, the molecule fumarate, for example, has got an anti crossing 30:08.499 --> 30:15.499 field and it happens around 400 nano Tesla. the definition of an anti crossing field is kind 30:15.499 --> 30:22.419 of where two states cross like that. And at the crossing point then you have this derivative 30:22.459 --> 30:29.338 of the frequency with respect to field is zero. So it means then that even if you have a small 30:29.339 --> 30:35.859 gradient then it's actually not going to broaden the line as much as it would for for T2 star. So 30:35.859 --> 30:42.818 for this molecule, it led to a very impressive, uh, decay constant. Um, 30:43.300 --> 30:47.219 and we're looking into that more. So there's certainly interesting things still to be explored 30:47.220 --> 30:53.539 in this ORF, even if we don't think about specifically about micro fabrication. Uh, now also 30:53.539 --> 30:59.379 interestingly, is that if we used a much more concentrated, uh, solution of fumarate, so we 30:59.620 --> 31:06.380 instead used a fully labeled version of the fumarate with carbon. This led to a much faster 31:06.380 --> 31:12.369 decay of that coherence. We think that's due to the fact that there are these distant dipolar 31:12.370 --> 31:18.529 field type effects which, which, which are big enough then that they actually do cause 31:18.530 --> 31:24.729 decoherence at the vapor. Okay. Anyway, that's something that's that's still being followed up. 31:24.930 --> 31:31.249 So I wonder then and I think we're in time. Um, 31:31.489 --> 31:38.448 by summarizing that, um, well, we kind of work in two 31:38.449 --> 31:44.369 areas. We work with magnetic with nuclear magnetic resonance, but we also work with magnetic sensors 31:44.929 --> 31:49.848 based on alkali vapors, which also destined for applications which have nothing to do with NMR. 31:49.890 --> 31:54.489 But there's a there's a lot of crosstalk and overlap in this area, and there's a lot of 31:54.490 --> 32:01.169 potential for miniaturize this and making, um, say, screening devices or relaxed symmetry 32:01.170 --> 32:07.839 devices at low field. Um, so we're still very much in the early stages of that. We don't 32:07.839 --> 32:14.679 study say systems, which are of, uh, super impressive chemical 32:14.680 --> 32:20.559 nature, but but it's going on. Um, and I also mentioned that there are there are going to be 32:20.560 --> 32:25.879 developments in the future that lead to these sensors becoming much more popular. The fact that 32:25.960 --> 32:31.159 MEMS technologies make these cheaper, and that there are other things in the pipeline as well, 32:31.200 --> 32:36.238 like there are some companies producing magnetometers which are not based on rubidium 32:36.239 --> 32:41.399 vapors, but based on other species like like helium. Okay. So then I just thank the people who 32:41.400 --> 32:47.239 founded the work, the collaborators and the rest of the team and yourselves. Thank you. 32:54.079 --> 32:59.519 Yeah. Thank you very much Michael for this very interesting and nice talk. Um, yes. 33:03.910 --> 33:10.469 Oh, thank you so much, Michael, for this fantastic talk. It's, uh, you always do magic with very cheap 33:10.470 --> 33:17.069 things. Very inspiring. Um, I have two quick, very quick, uh, comments and one 33:17.069 --> 33:23.509 question. Uh, the comments are that, um, your, um, uh, very narrow resonances near anti 33:23.510 --> 33:30.189 crossing. Uh, similar phenomena also happen in NV diamond. Uh, go by the name Morpheus or something 33:30.190 --> 33:37.109 like that, if I remember correctly. And they were studied, uh, for instance, uh, by many people, but 33:37.149 --> 33:44.149 uh, by Claudia Aviles at um in the Pines group some years ago. Um, and, um, 33:44.630 --> 33:51.429 the other comment is, uh, there is a large number of groups and companies in China that are 33:51.430 --> 33:57.948 into small magnetometers and, uh, in particular our former student Enam who 33:57.949 --> 34:04.859 started a company and they sell now, Um, uh, magnetometer that's very similar 34:04.860 --> 34:11.860 to Q spin, uh, in, uh, footprint. Uh, it is actually, uh, specifically optimized for 34:11.979 --> 34:18.019 ZULF NMR because its surface temperature is ten degrees lower and the cost is over another 34:18.020 --> 34:24.339 magnitude lower than the Q spin. So there's a lot happening. And it's by far not the 34:24.340 --> 34:30.939 only, um, company who does it. Okay. Those are my I will I just respond to that by saying, 34:30.939 --> 34:37.739 good., thanks for making me aware of those. I think that it's this, um, storyline that 34:37.740 --> 34:42.580 I've, that I've outlined here is, is a natural one. So I'm glad that other people are thinking of the 34:42.580 --> 34:49.379 same area and, and making progress in it. Yeah. And my quick question is you you showed that, uh, when 34:49.379 --> 34:54.739 you watch the movie of magnetization, which is cute, where, uh, there are some wiggles. Can you 34:54.740 --> 35:01.539 explain what the wiggles that you even modeled it, but you didn't explain where they come from? Uh, 35:01.739 --> 35:04.979 yeah. Shall I try and pull it up? I don't know how far back it was. 35:09.779 --> 35:16.459 Was it this one on this one? Um, okay. So it's 35:16.459 --> 35:23.379 perhaps easier to to understand the nature of the wiggles in the top. The top, uh, data set. Um, 35:23.379 --> 35:29.179 you can think of this like an adiabatic, uh, transition where you go from the north pole of 35:29.179 --> 35:34.459 the block sphere to the south pole of the block sphere. So a perfectly smooth adiabatic transition 35:34.460 --> 35:40.179 would be like following a meridian like that. But in reality, it's not like that. You always have 35:40.179 --> 35:45.700 some kind of precession about the field. So it it does this kind of trajectory. Uh, and those are 35:45.700 --> 35:52.659 called cycloid arcs. And so if you just take a projection of that, like onto, I guess onto the 35:52.659 --> 35:58.648 Z axis, then it looks like that curve I see. So it's not, it's not just um, uh, the 35:59.090 --> 36:03.889 hyperpolarization process, but the transfer of polarization that you're observing here. Right? 36:03.929 --> 36:10.809 Indeed. And then the the example at the bottom, why do you see much bigger wiggles? Uh, this is a 36:10.809 --> 36:17.809 system which is, uh, has lower magnetic equivalent symmetry of the hydrogens. So one 36:17.810 --> 36:24.009 way of saying that is the carbon makes the two protons look more magnetically and equivalent. Uh, 36:24.009 --> 36:30.810 and that leads to more polarization oscillation. A quick suggestion if you feed this 36:30.810 --> 36:37.129 back into the system, you can make a oscillator with this. And I will talk about it in my talk at 36:37.129 --> 36:43.889 the very end. Okay. Yeah. Yeah. Yeah. 36:43.930 --> 36:50.449 Great job. Thank you very much. I had an immediate question that came up was, uh, 36:50.570 --> 36:57.079 you make these vapor cells and, um, a very elegant process. How? How? 36:57.200 --> 37:03.280 What is the smallest size? You're allowed to make it so. In other words, you know, 37:04.439 --> 37:10.919 if I. If I told you I have a, you know, an Intel process with seven nanometers, we would be able to 37:10.959 --> 37:17.199 pattern something at seven nanometers. It would work, probably. But what would be the the what? The 37:17.199 --> 37:24.040 what? Yeah. What's the size limit? The how small these cells can. You can still make use of it. Yeah. 37:24.040 --> 37:30.560 There wasn't enough time to go into it. But a critical factor, um, in the sensitivity of these 37:30.560 --> 37:35.800 devices is then the relaxation time of the spins. And this, of course, depends if you think of gas 37:35.800 --> 37:42.799 phase at particles inside the cell, at some point when you make the cell 37:42.799 --> 37:49.319 smaller, you're going to run into wall collisions. So when the atoms collide with the walls then 37:49.319 --> 37:55.869 they will lose coherence. Um, you can do things like add certain coatings to the walls, which 37:55.870 --> 38:02.229 prevent relaxation. They provide elastic collisions, but there is a limit. And of course, as 38:02.229 --> 38:07.550 you shrink the size of the cell, you're also shrinking the the the number of atoms in the cell. 38:07.550 --> 38:14.428 So that's reducing the sensitivity. Yeah a follow up question on on on the cells, uh, how 38:14.429 --> 38:20.509 good the ceiling is of those cells. And do for example, do you notice, um, change of the, of the 38:20.510 --> 38:27.509 vapor volume over time. And how does that affect the performance? Um, that would 38:27.510 --> 38:33.869 help if I got a picture of one of the cells. So let's go back. Maybe to 38:33.909 --> 38:40.748 here. Something like that. Okay. This firstly, there's plenty of rubidium inside this cell. 38:41.310 --> 38:47.310 Um, the droplets here, uh, if you think if, if you operate at the cellar temperature, 38:47.990 --> 38:54.500 um, and you at some point, then the room, the the derivative rubidium leaves the cell 38:54.539 --> 39:01.299 by kind of permeating into the walls of the cell and of the glass. The glass firstly is 39:01.419 --> 39:08.059 covered with an alumina coating, which prevents permeation. Um, but additionally there's so much 39:08.060 --> 39:13.459 rubidium in the in the vapor cell that it would take a couple of years. I think it was ten years. 39:13.460 --> 39:19.659 We calculated for, for all of the rubidium to, to diffuse out. So there's so much rubidium in the 39:19.659 --> 39:26.419 cell because it needs a high nitrogen pressure for the, for the quenching, um, and for the buffer 39:26.419 --> 39:32.779 gas. Uh, and then because this is produced by azides. So rubidium is a bit like a 39:32.819 --> 39:39.378 byproduct. Um, you could, you could make these cells a slightly different way. So instead of you, 39:39.419 --> 39:46.419 instead of using so much azide, uh, which would create leftover rubidium. What you could do 39:46.419 --> 39:52.810 is do the bonding step under a higher back pressure. this is done in a commercial anodic 39:52.810 --> 39:57.489 bonding machine, which only goes up to a certain back pressure. I mean, you could put the whole 39:57.489 --> 40:02.929 thing in a pressurized room, but I think that would be that would be harder. Okay. 40:04.289 --> 40:05.649 Any further questions? Yeah. 40:11.330 --> 40:17.929 Yeah. Thanks, Michael. The quality requirements for the laser. You you 40:18.610 --> 40:25.249 went into that briefly that the laser is kind of a dominating element to the cost of the, uh, of the 40:25.249 --> 40:31.409 apparatus. So what are. Could you comment on what the quality requirements are for the laser 40:31.409 --> 40:35.849 radiation that's necessary for these measurements? Um. 40:38.209 --> 40:45.168 Okay, so these these are low power lasers, um, low power VCSEL lasers. And so 40:45.330 --> 40:50.679 low power, meaning around one milliwatts of infrared light. That's sufficient to get that 40:50.680 --> 40:57.159 level of ten femto Tesla sensitivity. Um, so these VCSEL can be produced 40:57.159 --> 41:03.359 in large batch quantities. So, you know, on a wafer scale, maybe you would have 41:03.360 --> 41:09.878 10,000 units on, on wafer, but there would of course be a yield 41:09.959 --> 41:15.519 factor. So there would be parts of the wafer which were lasing at the correct wavelength and some 41:15.520 --> 41:22.439 which were not lasing at the correct wavelength. The question or the issue really is scale. So if 41:22.439 --> 41:28.959 you had, um, if you went, if you went to one of the big laser manufacturers and said, I want to 41:28.960 --> 41:35.399 make a 7.95 nanometer laser for rubidium spectroscopy, um, with this 41:35.399 --> 41:42.158 recipe, they would say, great. Um, how many, how many millions of millions of units. So you can do 41:42.279 --> 41:48.709 to order. So for low production runs, it's not something which is very economical. And it's it's 41:48.709 --> 41:55.669 kind of difficult to do at the moment. Yes, yes. Yeah. I'm over 41:55.670 --> 42:00.269 here. So I have a very general question to the laser readout and the operational principle of 42:00.309 --> 42:04.709 the vapor cell. So the way I understood it, you shine a laser all the time, and then you look at 42:04.709 --> 42:10.628 the reduction in, in the, in the signal. So would you benefit from pulsing the lasers or you have 42:10.629 --> 42:16.549 it on initially. Then you let the system evolve in the dark. Yeah, exactly. This one. And then turn it 42:16.549 --> 42:23.269 on again. Or how much can you gain by this? Maybe this is actually the dynamic video you show. There 42:23.270 --> 42:29.229 are there are there are other types of schemes that one can use. So you could use, um, continuous 42:29.229 --> 42:36.029 pumping and slightly off resonance probing. Um, so but the cost of that is then you need two lasers 42:36.029 --> 42:42.149 at different frequencies. This is the one which is used. This method is the one which is used in all 42:42.149 --> 42:48.980 of these miniature MEG type magnetometers, because the fewest component is the simplest and 42:48.980 --> 42:55.978 smallest. The or easy to manufacture a small. So it's really just like this. And 42:55.979 --> 43:00.499 then you have to simply find a way of arranging those components small. It's not actually 43:00.500 --> 43:07.340 implemented like this. It's more implemented like an IQ modulation demodulation approach. 43:08.259 --> 43:15.139 So you're detecting the derivative around the peak here at zero field rather than uh, uh, 43:15.139 --> 43:20.139 rather than the absolute value of the transmission. Uh, but these are pretty robust 43:20.139 --> 43:25.099 things. So you can you can also make the electronics for this yourself out of Arduino. It's 43:25.100 --> 43:31.459 uh, that's not a factor of the cost of the devices. Is it a question I have. So with NV centers, uh, 43:31.500 --> 43:36.259 we can probe like you probe, but then we see the power broadening by the laser. So it's very 43:36.259 --> 43:42.299 beneficial to pump the laser, turn it off to the sensing, and turn the laser on again. So it's a 43:42.299 --> 43:49.129 similar type of letting the system evolve in the dark. Beneficial for wafer cells. Uh, it can 43:49.129 --> 43:56.129 be. So, um. Yeah. In this type of sensor, then if you increase the power level, there's a 43:56.129 --> 44:02.010 threshold where the power broadening reduces the sensitivity. It's that's not at the level that 44:02.010 --> 44:07.409 these laser VCSEL lasers can reach at the moment. But you could you could imagine an FID by 44:07.450 --> 44:14.209 magnetometer where you you pump the vapor to polarize it. And then you would 44:14.210 --> 44:19.529 turn off that pump and have a lower power probe. Yeah. So that's done quite often. 44:26.449 --> 44:33.329 Michael. Hello? Yeah. Michael. Thanks. Great talk. Um, I was wondering if you make these cells smaller, 44:33.370 --> 44:39.449 as you would if you use very many of them, then you should also need less laser power per cell. 44:39.929 --> 44:46.679 Right? Um. Or is that not right? Okay. So this is sort of 44:46.960 --> 44:53.079 related to the question that Jan asked. If you make the cell smaller, the cost is also that the 44:53.080 --> 44:59.879 number of sensing atoms decreases. So you want there is some kind of optimum if you want the if 44:59.879 --> 45:06.398 you want to maximize sensitivity, um, you would want to make your cell as big as, as 45:06.399 --> 45:13.279 tolerable. And also you would need to have a sufficient power of, uh, or you need to have a high 45:13.280 --> 45:19.199 enough temperature to use all of the available lights that you are shining on the vapor. 45:19.959 --> 45:26.719 Um, so it's a it's a multi parameter item. But but the reason to use many cells is to get a spatial 45:26.719 --> 45:32.759 resolution right from the, from the readout. Right. But these are separated by centimeter 45:32.760 --> 45:39.519 distances. Okay. But if you so if you go to millimeter distances could you then split the 45:39.520 --> 45:46.509 laser beam. Right. So that you use a few lasers to drive many such cells. It's possible. It it 45:46.509 --> 45:52.070 depends on the configuration. So you could have several vapor cells close together in a kind of 45:52.110 --> 45:58.070 little array. Um, that's not a good approach or that's not an approach for this zero field 45:58.070 --> 46:03.310 resonance, because you need to have a field modulation on each of the cells, and then you have 46:03.310 --> 46:10.309 crosstalk, and then it's going to become a mess. Yeah. But for if for cells which, uh, are far 46:10.310 --> 46:16.830 enough apart that there's no crosstalk,then, then that's an acceptable magnetometer strategy. But 46:16.830 --> 46:23.189 there are other, you know, like other types of strategy, like, uh, like FID. Thanks. 46:25.590 --> 46:32.549 Any further questions? Okay. If that's not 46:32.549 --> 46:35.469 the case, then let's thank Michael one more time. Thank you. 46:40.550 --> 46:40.790 Yeah.