How gliding marsupials got their ‘wings’

Lizzie Gibney

Welcome back to the Nature Podcast, this week: testing super-precise clocks out at sea…

Benjamin Thompson

…and how gliding marsupials started to soar. I’m Benjamin Thompson.

Lizzie Gibney

And I’m Lizzie Gibney.

<music>

Timekeeping is important — it’s about more than just making sure I get to my next meeting on time. Accurate timekeeping underlies a lot of research and navigation systems. This is why critical timekeeping is done with atomic clocks. These ‘tick’ as atoms switch states and are much more accurate than your wristwatch.

Many of these clocks’ ‘ticks’ are based on microwave radiation — these microwave atomic clocks are pretty robust and found all over the place. They’re at the heart of GPS satellites, and used in scientific studies to synchronise the capturing of high-resolution images of space. But they’re not the best timekeepers that exist. That title belongs to optical clocks, which tick using higher frequency, visible light. The problem with these, though, is they tend to be huge, fragile, lab-based contraptions.

But this week in Nature, a team describes an approach they’ve used to create a clock that combines the portability of a microwave clock with the precision of an optical one. To put it through its paces they tested it on a perilous journey at sea.

I called one of the team behind the timepiece, Jamil Abo-Shaeer, and he laid out the difference in precision between the two clock types.

Jamil Abo-Shaeer

So clocks that go into GPS, those are microwave atomic clocks. Those tick at maybe 9.2 gigahertz, 9 billion ticks per second. But the best clocks in the world — which are in the laboratories — they tick at optical frequencies, say 500 terahertz, which is 500 trillion ticks per second. So roughly speaking, the more ticks per second, the more precision. And so in moving from microwave to optical frequencies, we can do a lot better.

Lizzie Gibney

So your clock works at this more precise optical frequency. But your challenge was to try and take that out of the lab and make it portable. How did you go about doing that?

Jamil Abo-Shaeer

The atomic clocks in the laboratory, those are good now to a second in 30 billion years, which means that they're off by less than a second in the age of the universe. But the clocks don't operate for 30 billion years, they don't operate for a year, they operate for a day and so incredibly precise, but incredibly fragile devices. And a lot of it has to do with the fact that they're operating at these optical frequencies, which require very finicky lasers. And so part of what we did is instead of choosing say, the atoms that makes the best clock, we built our system around the most reliable lasers, we tried to simplify these clocks. So still making use of these optical transitions, but in a much more simpler, scaled down version.

Lizzie Gibney

And what is the atom or atoms that you chose?

Jamil Abo-Shaeer

So we actually choose molecular iodine, two atoms together, and that has some simple advantages, because of its insensitivity to environmental factors. So, it is less temperature sensitive, less sensitive to magnetic fields, things like that and so that's what we've settled on. The other piece of that is that we're able to make use of very robust laser technology.

Lizzie Gibney

So just how big is this clock, then? You mentioned it's compact, and how does its precision compared to the kind of clocks which live in the basement of a lab constantly monitored by postdocs?

Jamil Abo-Shaeer

So those ones in the lab, those are the most precise measurement devices on the planet: second in 30 billion years. We'd be very happy with a second in 30 million years and that's kind of what we are pushing towards. So not as good but still pretty ambitious in terms of the performance. So the clock goes into a standard server rack, and the size of that would be maybe a small microwave.

Lizzie Gibney

So this is not the best clock in the world but is it the best clock that can be used out in the world?

Jamil Abo-Shaeer

It's the best mobile clock in the world. It's the best clock that can operate under motion, least to the extent to which we know that. So the laboratory environment, the clocks don't move, the temperatures are usually regulated to a 10th of a kelvin, and the humidity is regulated to a percent. They typically operate on vibration isolated tables, and that's how they can reach these incredible levels of precision. From the beginning, we designed our clock to be a little bit more rough and ready. So it can do great in those laboratory environments. But when we took this clock out into the world, we saw little degradation in its performance in a mobile environment. In this case, we had it operate on a ship for three weeks.

Lizzie Gibney

That sounds like quite a stress test for the clock. In fact, it was a suite of three clocks, I think, on a navy ship.

Jamil Abo-Shaeer

That's correct. We went on a New Zealand navy ship. One of the challenges of having a very good clock is if you want to measure it and characterise it, you need another good clock to measure it against, you can't just measure it against your wristwatch. And so we have to bring identical clocks aboard. So these clocks operated in basically a glorified shipping container. There was a portable AC unit put it into the box. Which is great, it controls the temperature, but it was actually blowing right on our clocks and so the temperature was changing by as much as 7 to 10 degrees Celsius. Then there's the standard ship motion, we've all gone out on ships and gotten seasick. No different here, our guys had to get their sea legs. The ship itself can go up and down by almost 10 metres, the pressure changes out at sea significantly. So you kind of in a measurement like this wrap in all these real world things, but you make this technology for the real world and you can start really doing these kinds of interesting types of experiments.

Lizzie Gibney

And what were the results of that real world experiment? Like how well did the clocks perform?

Jamil Abo-Shaeer

In terms of mobile clocks, or say commercial clocks, the microwave clocks that are out in the world, they outperformed all those clocks. They're very similar in performance to Active Hydrogen Masers and so they are another type of atomic clock that is found in GPS base stations, which play a central role in GPS. But those are very finicky, they're too large and too sensitive to go into GPS satellites. They're typically about 300 litres — 10 times the size of our clock. But then those clocks also typically operate in an environmental chamber, that's 1,000 litres. So to be able to get that performance out of these clocks at sea, is very promising.

Lizzie Gibney

So we now have these very portable, very robust, and very precise clocks. How do you hope we'll be able to use them?

Jamil Abo-Shaeer

I think there's two things, we come from a scientific background and to be able to build metrology tools or scientific instrumentation that can aid in things like the discovery of black holes or quantum computing is exciting for us. But I think what's even more exciting is for us to contribute to current generations of Global Navigation Satellite Systems and future generations as well.

Lizzie Gibney

What kind of improvement would you be able to make to GPS if you could use these optical clocks rather than the atomic clocks they currently have in them?

Jamil Abo-Shaeer

So GPS is currently good to metre scales, its operating at kind of nanosecond levels, we would like to operate at picosecond levels that could give you precision down to centimetres. And that's important if you're in an autonomous vehicle, you want to be able to stay in your lane, and only being good to a couple metres isn't good enough. If you're landing a plane autonomously, you'd like to know what your altitude is to centimetres, not to metres, so you have a smoother landing. There are a lot of challenges in GPS, it's not just put a better clock in there's other work to be done.

Lizzie Gibney

Would you have to shrink down the clocks a little more to fit on a GPS satellite?

Jamil Abo-Shaeer

In order to operate in space, it's an even more challenging environment. We see going land, sea, air, to space in terms of difficulty. And so space is challenging because we don't get to touch the clock for 15 years, it has to operate in a challenging radiation environment, challenging thermal environment, it has to survive launch. Ultimately, you know, we are working toward that we're building a five-litre version of this clock, and that's about the right size of the current GPS clocks that are four litres. So we have to work toward lower power, we have to work toward better autonomy, and smaller size, but we think we'll be there.

Lizzie Gibney

And talking about space. NASA recently called for scientists to find a way to define lunar time, time on the Moon. And that's going to need an atomic clock. Would you like it to be one of your atomic clocks?

Jamil Abo-Shaeer

Absolutely. We're not the only company in the world working on this, but we have very ambitious goals. And if NASA is going to put a clock on the Moon yeah, we would love it to be our clock.

Lizzie Gibney

That was Jamil Abo-Shaeer from Vector Atomic in the US. To read his paper, look out for a link in the show notes.

Benjamin Thompson

Coming up, the genetic changes that allowed a tiny marsupial to fall with style. Right now, though, it’s time for the Research Highlights with Dan Fox.

<music>

Dan Fox

When some people break up with their ex, they move on by burning reminders of their relationship. But one of the kingdoms of the ancient Maya seem to have taken this ‘out with the old’ idea much further, as there's evidence of ritual burning of a previous royal family's remains when the regime changed. Researchers discovered burnt human remains and artefacts in a pyramid at a Maya site in Guatemala. The artefacts include a precious-stone mask of a type typically placed in royal tombs, along with other items that suggest that the remains belonged to Maya royalty. However, the archaeologists estimate that one of the individuals died up to a century before the fire, implying that former members of the royal family were moved out of their original tomb to be burned as a new regime was ushered in. The team behind this paper say that the fire was probably a public event, marking the beginning of the reign of a leader named Papmalil. If that is inflamed your curiosity, then you can read that paper in full in Antiquity.

<music>

An observatory at the South Pole has made the first solid detection of a type of elementary particle called the tau neutrino from outer space. Near massless, neutrinos are notoriously difficult to detect, and the tau neutrino has proven the most elusive yet: only being directly observed in our lab for the first time in 2002. One way to spot them is with detectors encased in ice, like those of the IceCube Neutrino Observatory embedded throughout a cubic kilometre of Antarctic ice sheet. Researchers looking at IceCube data from 2011 to 2020, used machine learning to help distinguish between the signals of tau, electron and muon neutrinos, eventually finding seven interactions that had a high probability of being produced by high-energy tau neutrinos. If that research sounds cool to you, you can read it in Physical Review Letters.

<music>

Benjamin Thompson

Sometimes, when faced with similar challenges, unrelated animals can evolve strikingly similar solutions. This is known as convergent evolution. And this sort of evolution can often leave scientists wondering whether there could be some genetic explanation for evolution turning up the same solutions over and over again. And that is the focus of our next story. Reporter Nick Petrić Howe called up Ricardo Mallarino who has a paper out in Nature this week focussing on the genetic basis of a flap of skin in marsupials that allows them to escape predators. Here’s Nick.

Nick Petrić Howe

So you've got paper coming out in Nature, that concerns something and I'm not quite sure how to pronounce it. Is it pata-jee-um and pata-gee-um?

Ricardo Mallarino

Pat-aj-um.

Nick Petrić Howe

Patagium, all right, I was completely off.

Ricardo Mallarino

We can refer to it as gliding membrane as well. Yeah, this is a very unique structure that has evolved a number of times in mammals and also outside of mammals. But yeah, it's a very interesting adaptation that allows these mammals to glide, which is a very efficient way of moving around, you know, escaping predators, accessing novel food sources and such. So it's a very critical adaptation and we wanted to understand how it formed from the molecular and genetic perspective.

Nick Petrić Howe

And, you know, what was it that drew you to looking at this particular structure?

Ricardo Mallarino

There's a number of reasons. And so in these marsupial's that I study, the gliding membrane forms when they're already outside of the mother, and inside the maternal pouch, because marsupials are very underdeveloped when they're born. And so you know, they're born, they will climb inside the maternal pouch, and they will complete their development inside the maternal pouch. And so that's very interesting, because it gives us a window into these processes that we couldn't see in mammals that stay inside the female for a very long time, like eutherian mammals. And so what we can do is, you know, temporarily, maybe take the joey out of the pouch, observe this tissue, study it as it expands and grows, do some kind of experiment. And then we put the joey back into the pouch, and the organism continues developing just fine. So that's one of the main reasons why I chose to study the structure in marsupials.

The other main reason is that this same structure has evolved independently in three closely related species of marsupials. And so that means that we can use things like comparative genomics, we can take the genomes of all these other species, the ones that have the gliding membrane, and the ones that don't have the gliding membrane, we can compare them. And then we can figure out if there are certain features of species that have the gliding membrane that are shared among them, and that are not shared with species that don't have the gliding membrane. And because they're very closely related, we can make very meaningful comparisons.

Nick Petrić Howe

Right. And in this study, you're trying to, I guess, understand the sort of molecular mechanisms, the genetics that underlie how they have these sort of skin membranes and how they develop. Where did you start with this study?

Ricardo Mallarino

So you know, there's a number of things that we had to do in order to be able to study this. One, is we needed to have access to a lot of these tissues from the species that we ended up generating genome sequences for and so for that we work very closely with museums in Australia that have these very large collections and so they were very generous with us. And we were able to get all these specimens and so we started sequencing all these genomes, right. That's sort of like the first thing that you need to do, because we're going to compare these genomes from gliding species to the genomes from non-gliding species.

Nick Petrić Howe

So you had these genomes from marsupials that had these gliding membranes and ones that didn't. Did you find any features within their genomes that sort of indicated why some did and some didn't?

Ricardo Mallarino

Yeah, so you can use various computational and statistical approaches to estimate the rate at which each of these genomes is evolving. And then we can find specific regions of the genomes that are evolving at faster rates than the average rate for a given genome, we find that all glider species seem to be evolving at a very high-rate in a particular region, nearby a gene called Emx2, indicating that this gene is probably very important for a particular trait that is shared only among the species and not among the non-glider species.

Nick Petrić Howe

Well, let's talk about some of the experiments you did on this gene. So you have this Emx2 gene, and you said that you had like the marsupials that you could do experiments with them, were you able to sort of like change the expression of this gene, like change how much of this gene there was, and that sort of thing and see what effects it had on the gliders?

Ricardo Mallarino

This is when the, you know, life history of these organisms came in very handy for us, because as I was describing earlier, the joeys are developing inside the maternal pouch. And so we can temporarily, you know, put the female to sleep with anaesthesia. And then we can take the joey out of the pouch, and then, you know, try to manipulate the expression of that gene, the activity of that gene, and then we put the joey back in the pouch, the joey is completely fine, continues living just fine. And then we can assess the effects of that. And then when we did, we found that we had managed to actually decrease the activity of that gene. And that led to a drastic change in the morphology of this tissue. And so we got gliding membranes that were much shorter. And so this is how we conclude that this gene is probably a key element that plays a very important role in shaping this very important structure.

Nick Petrić Howe

Do you think this paper sort of answers your original question of the sort of what is behind this novel characteristic that they have?

Ricardo Mallarino

You know, unfortunately, every time you complete one of these studies, there are more questions than answers. And I think that's the fun part about science. I think it opens a whole new set of possibilities and I think now we have a lot of other questions, right? For example, you know, is there something that is regulating Emx2? Right, like why Emx2 and not other genes? What about you know, outside of gliding marsupials, right? There are other species that have evolved gliding membranes outside of marsupials, right? There's this Colugo, which is this primate species in Southeast Asia, there are flying squirrels, obviously, which are a little bit more well known. And even outside of mammals, I have a postdoc in my lab that is now studying gliding geckos, right? These are, these are reptiles that have also evolved these membranes, we think they're very different in their composition. But we think we have now some preliminary evidence suggesting that they form through the same molecular mechanisms, at least initially, right? We don't know what happens down the road. But initially, many of these events are triggering the formation of the structure. And so what we are beginning to think, is that there is this genetic programme that is deeply conserved across all species. And all you need to do is like, tweak it a little bit, and then everything is in place to make one of these gliding membranes.

Nick Petrić Howe

Do you think there are sort of broader implications of your work?

Ricardo Mallarino

I think so. Because if you think about it, you know, 99% of studies that are out there are done in so called, you know, traditional model species. But you know, life is extremely diverse. And so we're just sampling a tiny, tiny, tiny tip of the tree of life. And so I think we're missing out on a number of different things that can be going on. But you know, nowadays, because of the powerful technologies that are out there that are more accessible in terms of cost and in terms of the technical aspect. Now, I think it's a great time to, you know, diversify from studying these sort of like five or six species and then see what's out there.

Benjamin Thompson

That was Ricardo Mallarino, from Princeton University in the US. For more on that story, check out the show notes for some links.

Lizzie Gibney

Finally, on the show, it's time for the Briefing Chat, where we discuss a couple of articles that have been featured in the Nature Briefing. I believe, I'm going to start today. So as ever with me, AI is the name of the game.

Benjamin Thompson

Right.

Lizzie Gibney

So this is a story from Quanta Magazine, and it's on a phenomenon called grokking. Have you ever heard of grokking?

Benjamin Thompson

I mean, as someone who's played a lot of RPGs. I imagine that that's in there somewhere, but not entirely familiar, no.

Lizzie Gibney

So grokking is a term that some researchers coined and it comes from science fiction, and it was an author's term for understanding a phenomenon so well you almost become a part of it.

Benjamin Thompson

Right, so how does this relate to an AI story then?

Lizzie Gibney

So grokking is when an algorithm gets good really, really quickly in a way that they're still trying to struggle to understand. So what tends to happen is you train a network and it gets better and better your training on the data and you give it unseen data, and it starts getting better and better and better on answering questions on that unseen data. But at some point, you hit overtraining, where it's kind of memorised all the data it has, it's really good on that, gets 100% on that, but starts to go down in performance with this unseen data.

Benjamin Thompson

Which is not what you'd necessarily expect, I suppose.

Lizzie Gibney

Well, that's not what you want. So this is what they try and avoid, they try and avoid overtraining. But what happens — as a team discovered when they accidentally left their model training, when one of them went on holiday — is that if you overtrain to such a large extent, this phenomenon called grokking can happen, where then actually, the algorithm gets incredibly good, and started answering 100% of the time correctly, on the questions it was posed on unseen data.

Benjamin Thompson

So rather than this fall off then, it's inverted commas “brain” gets significantly larger.

Lizzie Gibney

Exactly. So what is exciting about this is, in this case, we're talking about a teeny, tiny, very specialised neural network. But if you could try and make algorithms that are much bigger, that are much more useful have this phenomenon, that would be absolutely incredible. So I should probably tell you a little bit about the actual model that they were using here it’s a maths one. So it was getting it to calculate what A plus B is, A plus B equals C, but not with conventional maths with kind of like clock maths. I don't know if you remember doing this at school where 11 plus 2 is not 13 but it's 1 because it's like, you’ve gone all the way around the clock–

Benjamin Thompson

–Ok, yeah yeah–

Lizzie Gibney

–and you come out on the other side. So it's what you thought trying to get it to answer is that correctly. Except in this case, they didn't use 12 they used 97.

Benjamin Thompson

Because reasons.

Lizzie Gibney

Because they wanted to make it that little bit harder. And so what they think was happening is, as they put all of this data in and it was learning, it kind of went from memorising to generalising. So when researchers tried to pull apart this particular algorithm, but also lots of other similar ones that they built, that, again, all very, very simple and the same kind of questions. When they did that, they found that what the algorithm seemed to be doing was finding a solution, it wasn't the kind of solution that we might have found, it was using very, very complex geometry, some very difficult mathematical functions, but it was coming up with the answer that way. And so what the researchers who this journalist talked to think is going on is that you've got these two processes happening in parallel. So the neural network wants to be efficient. So it just builds up effectively this memory, it knows the right answers to give. But simultaneously underneath, it's doing this kind of generalisation until suddenly, at one point, it finds that the general solution works much better than the memorised solution, and then employs that one instead. And it's kind of got this formula that it's come up with actually works all of the time.

Benjamin Thompson

I mean, we talked on the podcast a while ago about an AI that was trying to learn the kind of basic rules of maths. And this seems like it's in the similar ballpark. Presumably, for people who are interested in this, this is quite useful like you can suddenly get a leap in performance for your AI.

Lizzie Gibney

So as I mentioned at the beginning, this is very small, very specialised and we’ve really got to hammer that home, because we don't know if it will apply when you start using bigger networks, bigger algorithms. But the idea is, if you could do that, you would get a huge increase in performance. But I mean, there's a cost to it as well, imagine if you went around trying to over train all of your models, that would be an enormous amount, potentially, of time and effort wasted. But what I think this really speaks to is efforts to pull apart the black box of AI, which we've discussed before, and trying to figure out what might actually be going on, what might be behind these kinds of phenomena when they happen, because in this case, they also found that if you apply certain restrictions to the model, you can push it towards that generalised solution. So if you kind of limit how complex its mapping is inside, you make it stop relying on its knowledge and move towards figuring out a one-stop shop solution, which in my head is a bit like going from, people are gonna find this offensive, but like chemistry, where you have to just learn a lot of stuff, to like physics, where you just got to learn some principles, and it all falls out of that.

Benjamin Thompson

That’s the sound of a lot of chemists turning off the podcast there Lizzie. Anyway, let's move on in this week's Briefing Chat, and I've got a story that's got, well it’s got a little bit of bling to it, okay. And it's a story that I read about in Nature, and it's based on a paper in Nature Synthesis, okay. And it's about the gilded cousin of graphene. Now, I know that you have written a lot about graphene over the years, but this is about ‘goldene’.

Lizzie Gibney

Well, things that end in ‘ene’ tend to be single layers, like one atom thick layers of different materials. So this is atom-thick gold?

Benjamin Thompson

Absolutely right. It is yeah so g-o-l-d-e-n-e, goldene. Which my brain keeps on reading as Goldeneye. And apparently there's a Pokémon with a similar name, but it's not that, it has nothing to do with that. It's to do with 2D layers of gold atoms. Now, these were long hypothesised to have been able to actually have been made but it's been a bit of a struggle doing it. And now this appears to be the first free-standing goldene, so unattached to anything else, according to the researchers. There was a previous report of it being done but that is somewhat contested. But it's been quite the journey to get there and they came about it in a bit of a weird way.

Lizzie Gibney

So graphene was discovered by tearing off effectively a layer with sellotape. I'm guessing this was a lot than that?

Benjamin Thompson

I mean, there was some fortuity to it I have to say. So researchers have made 2D layers of some metals like tin and lead, but they've been stuck to other things. It's really hard to make 2D sheets of metals, right, because the atoms tend to clump together. And as I say, there's a little bit of luck here. So what the researchers wanted to do, they started with a 2D layer of silicon atoms, and they sandwiched that between two sheets of titanium carbide, right. And the idea was nothing to do with gold at this point, they were trying to make this component for some research they were doing right, but they wanted to coat this component in gold. So they did so and it turned out that when they heated it up the gold and the silicon kind of swapped places. So they had sort of a very thin 2D layer of these gold atoms sandwiched between this titanium carbide. And it was stuck there for several years because they couldn't work out how to get out.

Lizzie Gibney

Oh wow–

Benjamin Thompson

–yeah.

Lizzie Gibney

How did they end up getting out then?

Benjamin Thompson

Well, they actually used a process that's been used in Japanese knifemaking for over a century.

Lizzie Gibney

Oh, wow.

Benjamin Thompson

Yeah, so they use something called Murakami's reagent. Now, this is used to etch away carbon during knifemaking. It's been used this way, say for a hundred=odd years. And they use this to remove the outer layers of the sandwich. And they do some other chemical sort of techniques as well. But what came out the other end was goldene, this 2D material made of just gold atoms.

Lizzie Gibney

So aside from being the most expensive thing you're very likely to lose. What is this useful for?

Benjamin Thompson

Well, that's a great question. And the answer at the moment is maybe a lot of things, but we don't really know, right. So as you know, graphene has some remarkable properties, right. It seems that changing something from a 3D orientation to a 2D orientation, allows it to do some quite odd things. And that might be the case with goldene as well. Now, gold is used a lot in a lot of different industrial processes and gold nanoparticles have shown some promises like catalysts and reactions and stuff. So it's thought that maybe this kind of very thin layer of gold could be used in things like splitting water to make hydrogen, potentially could be used as a semiconductor because normally gold's a conductor, but this seems to be changed that property, I think people are quite excited about it as well. But for the time being, really, I think that researchers want to work out how to improve how they make it. And one researcher in the article was quoted as saying that you know they need to be careful that there's no other atoms like atoms of iron that are contaminated from this etching process. But yeah, it seems to be the thinnest gold-leaf that exists on Earth.

Lizzie Gibney

That's amazing. And I know that a lot of the time now 2D materials, researchers are trying to just build different kinds of sandwiches, some of them, you know, many layers high that have a whole different range of properties. So I guess now you've just got another material, you can shove in that sandwich and maybe create whole new phenomena doing that.

Benjamin Thompson

Oh 100%. And researchers they say it's early days, but they really want to see whether this method can be used to make other monolayers as well, right other important metals like iridium, platinum, and palladium. So these are important catalysts too and it could be as you say like this really opens the door to multi-material, really flat sandwiches that have interesting chemical properties.

Lizzie Gibney

Fantastic. Well, thank you for that, Ben. Listeners, for more on those stories check out the show notes for some links, and for a link of where you can sign up to the various Nature Briefings to get more stories like them delivered straight to your inbox.

Benjamin Thompson

And that's all for this week. But before we go, just time to say that you can reach out to us on X, we're @NaturePodcast or on email podcast@nature.com. I'm Benjamin Thompson.

Lizzie Gibney

And I’m Lizzie Gibney. Thanks for listening.