Hello, and welcome to the Physics World weekly podcast.

I'm Hamish Johnston.

In this episode, I'm in conversation with Antonio Rossi, who's a researcher in two d materials engineering at the Italian Institute of Technology in Genoa.

We explore how two d materials such as graphene are finding a wide range of applications, including in fundamental science, quantum technologies, and industrial coatings.

Much of Rossi's research focuses on tungsten disulfide and hexagonal boron nitride, which he explains are two d materials with great potential for electronics and optoelectronics.

We also chat about the use of artificial neural networks and artificial intelligence to optimize the synthesis of two d materials.

Before we get to that conversation, I'd like to introduce you to a new and quick way of discovering cutting edge science.

IOP Publishing's new Progress In series, research highlights website offers quick and accessible summaries of top papers from leading journals like Reports on Progress in Physics and Progress in Energy.

Whether you're short on time or just want the essentials, these highlights help you expand your knowledge on leading topics fast.

Perfect for busy researchers, curious minds, and anyone who wants to stay informed without the deep dive.

To start reading, just type progress in series research highlights into your favorite search engine, or follow the link in the notes for this podcast.

Hi, Antonio.

Welcome to the podcast.

Hi.

It's a pleasure to be here.

So, Antonio, can we start with the basics?

What are two d materials, and why are they interesting in terms of science and technology?

Yes.

So two d materials are, like, classical materials and arrangement of atoms.

The the the main difference with respect to, classical three d three materials is that they extend over a sheet that is, three up to three atoms thick, which makes them extremely different in terms of properties that they display.

And, they they go through what's called quantum confinement thanks to this, low dimensionality, which is great because usually atoms, they tend to be in the three d world.

Right?

They tend to be, x y z orbital, but then they are confined in a x y space.

And packing them together, basically, this, offers the opportunity to explore a different physics with respect to, normal Trinity compounds.

I see.

And and graphene is perhaps the most famous two d material.

In fact, I I I had an encounter with graphene over the weekend.

My wife and I are looking at buying a new car.

And we went to the dealership, and they were offering a special graphene coating on the car.

I sort of had my doubts as to whether it was actually graphene, or or something something else.

But, you know, I think a lot of listeners will be familiar with graphene.

What what is the current state of play with graphene?

Is it being used in commercial technologies?

Indeed, were they not joking at the car dealership?

Were they putting it on cars?

So they were not joking up to a point.

There's been a huge debate lately about what graphene actually is and what these components are made of because they it's easy to say graphene, then when you go look into that, it's not exactly graphene.

Graphene is mostly, let's say, one single layer of carbon atoms arranged in a honeycomb lattice.

Right?

And that's the definition of graphene.

But then people start playing with the having, graphene oxide that is a different structure as an oxygen in the in the compound.

Reduce graphene oxide that was before oxidized and then reduced, and then it's very defective, which is not graphic because it's missing some, some carbon atoms from from the from the lattice from the lattice structure.

Yet, is it that different?

Can we just, like, graphing based material, all of that?

That I think it's fair to say that.

And the dealership yeah.

Then they will probably offering you, like, this new coating that also I don't I'm not sure if I'm allowed to say company's name, but, even our our, Airbus is, is developing, for for airplane technology.

And so it's reasonable that you have, like, a composite material that implements some graphene form of, these, bigger umbrella of graphene based materials.

In that sense, there's been a lot of development towards market.

Like, let's say the the the first years of the what's called graphene flagship, the big European initiative to study graphene and two d materials, has been prompted by the need to understand the property of two d two d materials.

But later on, people were like, okay.

Now we have a clear picture of what's going on, at least, like, let's say, 80%.

Let's try to go to market.

Let's try to make a product out of it.

And so there are different sectors that you can think of, from composite material to electronics to, batteries.

And now the market the graphic market is around 350,000,000, 380,000,000, US dollars.

And but it's projected to reach the 1,500,000,000.0 by the 2027.

So it is being used.

It's implemented.

There's a there are a lot of, like, initiatives in that sense.

For instance, like, one great example that, came out recently is, like, this new graphene based paint that you can use to paint your house walls with.

There is also a heating source.

So, basically, you are cutting down, the power consumption to heat up and to, to heat up your house.

So you just get the heat straight out of the paint.

And so this is like a great, great, use of, of graphene in my opinion.

But I know, like, many people ask, so is do you find, like, a real application to graphene?

Because as the conversation started many years ago, everybody was expecting new electronics.

Right?

Exactly.

Yeah.

They say graphene is the new silicon, and so everybody was waiting for that.

But there are many problems in that direction.

First of all, it does not show the zero one, signal the silicon produces.

And so that makes, like, the the the logic the computer logic a little bit hard to to parse.

And at the same time, the quality of the material that, people use to make science from the famous Scotch tape technique used to be much higher than the, the quality of the material that you can produce via other methods and scalable methods.

Now this gap has been closed, I would say.

Like, lately, like, you can see that chemical vapor deposition graphing is famous technique that, employs, copper as a subject to grow graphene, delivered, like, very high quality samples.

And you can see the quantum physics that was that was, at first just in the exfoliated material.

And so that that that gap is causing, but this is something that is really recent.

There are a number of scientists across the the globe, that are working into integrating graphene into silicon technology, and I think that's a safer route because you don't have to reinvent the whole supply chain.

You're just placing, like, a piece of, carbon, the graphene, into something that already exists that can make it better.

On the other hand, there are some properties that only graphene has, like its optical absorption.

They makes it they make it extremely suitable for optoelectronic application, for extra fast auto optoelectronic application.

And there are a number of startups that are, like, start, becoming famous in that sense, like, becoming, delivering to the market some products that can actually be used to detect and motivate flight, which I think it's the closest thing to having something entirely graphene based, that is rephrasing electronics, of course.

I see.

So so it's still we we can still call graphene a wonder material.

It's just that, maybe it's taken longer than expected for some of those applications to come to fruition.

Yes.

Yes.

I wanna say that we are a little bit biased.

Right?

Because we knew what silicon could do, and we could not see anything different.

So we wanna have silicon but made of graphene.

So it requires, like, a, let's say, a bigger leap to reinvent electronics, to reinvent logic.

And in that sense, that might require a little bit more time, but graphene is still, like, a perfect candidate in that sense if you ask me.

I see.

And, Antonio, the graphene is by no means the only two d material.

There are others, that have great scientific and technical interest.

You work on tungsten disulfide and hexagonal boron nitride.

Can you can you give us an introduction to those materials?

Absolutely.

I I'm really excited to talk about these two materials.

Taq's and sulfide first because it was my PhD project.

You know?

I started, like, synthesizing, tanks and sulfide.

Packs and sulfide is basically like graphene.

It's still like a two d materials, but instead of being made by one layer of material, it's made by three layers.

So the top and the bottom is sulfide, so sulfur sulfur layer.

It is sandwiching a tanks and atom layer.

So it's like a really thin sandwich.

And, the with respect to graphene, it displays that zero one characteristic that is typical of of silicon.

Okay?

And has also great, optical properties.

It absorbs light.

It emits light.

It it works really well in that sense.

And given that it's three atoms, and we're still talking about something that is a very limited number of atoms interacting with light.

So given that, it works extremely well.

And recently, TSMC in Taiwan delivered, like, one technology that was entirely based on a cousin of tanks and sulfide.

It was molybdenum sulfide.

Basically, the same properties just like different numbers, but same properties.

So that in that sense, like, if you if you think about, like, regular electronics, that's like a strong candidate to actually outperform sync silicon, in the two d material world and can do things that graphing cannot do at the same time.

I see.

Just to make things clear for, if our listeners aren't that familiar with the product properties of semiconductors, when you talk about zero one, you're talking about, the fact that you can make a a a a a transistor or a switch or something that can toggle between two distinct states.

And that sort of forms the basis of of all our computers.

Whereas with graphene, there there aren't the two distinct states if you make a transistor.

Is that right?

That's right.

That's right.

So let's say, yes, that you can with some tweaks.

You can do that in graphene as well.

But let's say that the zero and one, the the two bits that they use to process regular information nowadays in our computers, are not so well defined.

So somehow it's hard to understand whether you're looking at a zero or a one.

At least it's harder than it's in tanks and sulfide where those two states are very well separated.

And so you can actually it's like the the Morse code.

You know?

You can have a, like, a zero one sequence that is encoding a message, and you are sure that the message you're seeing is the right one.

I see.

And and what about hexagonal boron nitride?

Is that is that a a newer material for you that you've you've just started working on, or have you been working on that for a while as well?

So it depends because hexagonal boron nitride is the the elephant in the room in the in the two d materials business.

So we have seen in the past years a number of physical phenomena that are that are hosted by two d materials.

Okay?

Reason for in 2018, there was, like, the superconductivity that emerged when you have twisted layers of graphene.

There's nothing of that that you can see if you don't send with your between hexagonal and boron nitride.

Exagger and boron nitride, like, let's say you have your, your layer and you have hexagonal boron nitride on top and on bottom, and then you find the perfect environment for two d materials to to thrive.

Okay?

Now that's the caveat, though, because all of that is done with, exfoliated boron nitride.

So there's, one big group in Japan and professor, Watanabe and Taniguchi from the National Institute of Material Science that basically provided that boron nitride to the community to the whole community, which was great because it allowed us to really do a lot of exciting physics with technically a scotch tape.

Okay?

The this is like a very democratic way of doing science.

You don't you don't need big machinery to access exotic physics, exotic phenomena.

And but then what if you wanna bring that that that effects, those phenomena to the market, to the scalable application?

There's no scalable boron nitride.

Right?

So one of the, I used boron nitride Scotch tape voice.

Like, I've done some of my devices with, like, a escalated boron nitride.

But now I'm working on the I just started working on the synthesis of, boron nitride large scale.

And, that's being done something in in the past.

People have been working on scaling, their boron nitride production.

The quality is still not quite there yet.

Many, many attempts have been done.

There are some, like, reports of high quality, CBD, chemical vapor deposition boron nitride, but it's still not there.

And this is something that, I think it's like the the the the holy grail of the of two d materials at this point is to find the right let's call it the dielectric, the right substrate.

The three materials are ideally, like, the floating vacuum.

Right?

But they need to be put onto something.

And that's something has been always at boron nitride.

Exceptionally, other strategies have been adopted.

There are other flat substrates.

But if you really wanna have the pristine properties of any material, so far, the boron nitride is unmatched.

I see.

So that's so so you so so to create a practical device, you you have to be able to manufacture it in a similar way in in a chip, processing facility.

And I suppose the the the the problem that you have is that once you put your lovely two d material onto a substrate, it's not two d anymore.

So you have to find a substrate that will hold the two d material, but not hold it so tightly that it changes it into a three d material.

Is that is that a big challenge?

It is it is a a a big deal.

Yes.

And, so the fact that as as I said at the beginning, having a two d materials means that it is actually really sensitive to the surrounding, to its surroundings.

So if you place it on silicon oxide so transistor these days are made of, silicon and silicon oxide structures.

Silicon oxide is rough.

Like, not rough as we might think of, like mountains rough.

But it for for something that is one atom thick, it is rough.

And so they create this creates inhomogeneities that make the two d material they're still two d in that sense because it you will still have, like, their their properties being confined to the graphene layer if you're talking graphene, but it will be affected a lot by the substrate being so inhomogeneous.

Boronitride, on the other sense, is as flat as as graphene.

And if you have many main layers, it can protect it from the surrounding environment.

So it's basically unperturbed graph.

Now, again, depends what one wants to see because if we are interested in just like the transport properties of graphene, it still is an incredible conductor even on silicon oxide.

No problem.

But if you're interested in taking advantage of the superconductivity phase as we as we talked before for the twisted system, that might be much more challenging on a silicon oxide.

Ship.

I see.

And, my understanding is that a lot of your research is, looking at ways to to address this challenge, and you use artificial intelligence and other computational techniques to develop new two d materials and in such a way that they can be scaled up, for, I suppose, commercial applications.

Can you can you describe that work?

Absolutely.

So this is something that, I've been working on for the past couple of years.

Just to give context, we, developed this technique onto something that we already knew how to synthesize, but we knew we needed a proof of concept that the the approach was actually valuable.

So we no surprise, we synthetize some graph.

Something that people know, how to do for for a long for a long time.

And but then the same approach can be also applied to, boron nitride.

What am I talking about?

I'm talking about the use of, neural network in that sense.

AI is a really wide, domain.

Let's talk about neural network that can propose a recipe.

Okay?

They learn, how to propose recipes material.

You provide a feedback to the to the neural network.

And the neural network be as its parameter adjusted in with the method that is called Monte Carlo.

That is, let's say, a stochastic approach.

It it's a it's a method that has been using physics for fifty years now to find the minimum of some of some states.

Let's say it's optimization problem.

So you wanna find the the the most efficient, configuration of your neural network that provides the best, neural network configuration for the best recipe.

Yes.

So this is basically the the the the overall approach, and it works really well for the graphing synthesis.

We we we recently published that.

It's also an easy problem, though, because for the graphene, we use, silicon carbide, as a precursor for our graphene.

So, basically, how it works is that you take a piece of silicon carbide.

You heat it up in a furnace.

The silicon atoms, leave the surface.

The carbon atoms that are left behind form this carbon rich subs carbon, rich layer that, we call graphene.

And, it's just like a temperature profile, so it's an easy task for for a neural network because we will also know.

But, technically, this method that has been, developed by, doctor White Lamb at Berkeley National Lab is, can be applied for multiple parameters.

So you can tune temperature, mass flow, pressure, and we're trying to implement that for this synthesis of boron nitride.

Yes.

I see.

And and can you talk a bit more about, about your work in the lab?

What sort of experimental techniques that you use to create and characterize two d materials?

I mean, I'm guessing that you're you're using things like electron microscopes and and various surface science techniques.

Yes.

Exactly.

So as a matter of fact, the surface science is my background, and, I've been working on photomission spectroscopy for many years now.

Spectroscopy in general, but photomission spectroscopy is my core business.

So we have, an ultrasound for the mission spectroscopy, that is coupled to a X-ray photomission microscopy that is also coupled to a scanning tunneling microscopy and a low energy electron diffraction.

Many names, many names, but they they are basically You must have a huge vacuum chamber then.

We we do.

We do.

We do.

As a matter of fact, yes.

To be honest, we have three, four vacuum chambers.

They're all connected.

So that's a major plus for us because we need to prepare the sample once, and then we can can use all of those techniques all at once.

And, yeah, we can basically have my old supervisor used to say we we we are capable of getting the full Hamiltonian of the system.

So the electronic state, the structure, everything that is necessary to characterize a a material, which is great.

It's a little bit slow, though, because you need to go to ultra high vacuum.

We need to pump overnight.

So in those in that feedback loop that I was mentioning before, I'm actually using, Raman spectroscopy.

There is an optical spectroscopy technique, extremely famous for two d materials, which is extremely reliable in assessing the quality of the of the system you're looking at.

So in that in that loop where I need to give a feedback to their to the neural network, I use Raman spectroscopy just because it's fast.

So I can produce a sample scanning under Raman spectroscopy.

Am I happy with it?

Am I not happy with it?

They get a better result, a worse result?

I tell that in a way to the neural network, and the neural network adjust its parameters and produces a new recipe.

And so that that's that's extremely, powerful in that sense.

Of course, we need to post validate that because Raman can also hide some stuff.

So you need to make sure that they you're interpreting the signal, in a in a in a proper way.

So then we use the surface science characterization tool to have a full understanding of what if we're going in the right direction or not.

But this is something that happens afterwards.

Yes.

Okay.

And and so does that mean you're sort of operating in a loop in the sense that your AI system will say, try this, try this, and try that.

You will do it.

Feed that information back into the AI.

It will come up with a better, prediction of a material that you want, and you're sort of continuously improving and and feeding back into the machine learning system.

Is that is that how it works?

Yeah.

Correct.

Exactly.

So it starts with a random guess by the neural network, and, I test it, and the the loop begins.

And then I start, like, measuring and then provide this number, the score function to the to the to the system and then so on and so forth.

It is interesting because it, we say we don't give any prior knowledge to the system.

The the the main so lately, there's been a lot of, hype be be behind the neural net the AI for material science, especially be because of the autonomous labs that promise to deliver new materials, thousands of new materials, and new and since which is great, but it's also, like, heavy in terms of data.

So it needs to be trained on hundreds of thousands of published results to understand what can be done, what are what crystals can be synthesized, how they you can do that.

The thing is that if you're training something onto something that already exist, it's likely to produce something that already exists.

So to rediscover a material that people already know, and this is something that's been, reported lately as a as a, let's say, bottleneck of this approach.

Our approach is, it doesn't it doesn't go through that kind of training.

So the the neural network learns by doing.

So it's like a baby learning how to walk.

And, surprisingly, it does not require so many iteration to reach, like, a, a satisfying outcome.

And, but it's extremely light in terms of data consumption.

And, this also makes it a little bit more transferable to university, labs that are not, funded with that kind of resources and, of the autonomous labs, of course.

And, that makes it extremely appealing for for us with the scientists still in the loop, not outside.

I see.

Okay.

And and finally, Antonio, I wanted to ask you about quantum applications because 2025 is the international year of quantum science and technology.

And I mean, I suppose in many ways, two d materials are sort of ideal for tweaking the, you know, quantum properties and and discovering new and exciting and useful quantum properties.

Can you can you just talk a little bit about how two d two d materials are being used, to both explore, I suppose, quantum physics and also develop new technologies for, I don't know, quantum computing or quantum sensors.

Yes.

Absolutely.

So the quantum is a is a powerful word in that sense.

Right?

So you can say that everything is quantum.

But, when you talk about quantum technology, you really want to, narrow down your field to something that is really explainable by something that's been accessible to people after 1925.

Some something that it's really, really described by Schrodinger equation and everything that came after that.

There's a whole field of that.

The quantum sensing is one of these.

For instance, you can use two d materials like boron nitride and the defects, which is interesting because now you're not really interested in the boron nitride per se.

You're interested into the defects that the boron nitride can host, and they act as a, basically, a spin sensor, magnetic field sensor.

So you can basically detect the presence of a magnetic field very locally, very precisely onto a material that, as we said, is really sensitive to the external environment.

So it's really powerful in that sense.

And this is just like one of a one of a one example.

Of course, defects are present in any of those two d materials.

And some of the TMDs, transition metal that you're co organized, like tanks and sulfide that we discussed earlier, is also a system that can host interesting defects because they are, so called two level system.

So the two level system now if we go back to the to the electronics and how it works, it it sounds like zero and one.

Right?

It's two states, zero and one.

One thumb tells you that those zero one are not just like zero or one, but are zero and one.

So it's what's called the qubit.

That that that way of processing information that it's not easy to interpret because it goes beyond our common knowledge and our common sense of the physics around us.

So you need to get familiar with this concept of superposition of states that you can have multiple combination of zero ones.

And tags and so five, for instance, is one of those system that can host defects that display those two those two states and can be accessed.

It can be manipulated.

It can be, used to, process quantum information.

And then there are, like, another other, two d base quantum system, like, graphing based Josephson junction that they used to detect microwave.

So you have superconductors that is this new, state of matter.

Well, no.

It's been known for a while, but it's a, it's a state of matter where current can flow with no resistance into a material.

And you can create this junction where you place one piece of that material, connect it to graphene, and then connect it against with another superconductor.

This is a Josephson junction with the weak link.

It's called weak link.

This graphene in the middle.

What happens is that it's extremely sensitive to microwave radiation.

So exploiting this intricate, collective phenomena called superconductivity, you can actually detect microwave radiation in a extremely powerful and sensitive way.

And so this is something that is, like, unprecedented with respect to non quantum technologies.

So it's really promising.

And this is just to name a few.

But there's, something that I'm really excited about is the, what's called quantum, matter.

That is, these these new physics that arise when you place two different material, one on top of the other.

And when you have either the same material with the lattice, misorientation, so there's a twist angle between between the two system or the entire different lattices, so with different periodicity, super structure arises.

Okay?

So you can imagine like an ad like a crystal.

Again, it's a periodic arrangement of of of atoms.

Okay?

So you have, let's say, a carbon atom to every x angstrom.

Okay?

And that periodically.

Now if you place another, carbon lattice on top of this one and you twist it a little bit, you will see a more structure.

So a super periodicity, something that has a longer periodicity than, than your than your constituent crystals.

It may look like just an optical effect, but that's in fact, electronic potential that the electrons feel.

Okay?

So they are basically immersing this new environment that the new periodic environment that they explore.

Some of these system, are formed in a way that this molecule potential is essentially a quantum well.

So a well that can trap electronic states, excited states, any kind of states.

So you are making like an artificial lattice out of two real lattices.

But now it's way more tunable because you are not limited by the chemical bond that keep atoms together.

You can twist, the twist angle as much as you want and change its periodicity or combine different lattice together.

You have will have an entire new periodicity.

And so this opens up, like, a number of possibilities.

And one of these, I think I find very exciting is, for instance, the quantum polaridonic, external polaridonic, field that is when to to have to use, like, a catchphrase is when light becomes matter or when light matters, meaning that, you will have a strong interaction between a crystal and light, and then you can trap the state of light and translate the state of light into electronic state.

You can manipulate it, and then you can release it as a new form of light.

And this is like a very exciting way of, working with light encoding information and quantum information.

I see.

Yeah.

I mean, it is a I mean, I think the whole moire thing is incredible because, you know, as you say, you can just with a little twist, you can create a custom lattice almost with, you know, the separation that you want to match, I don't know, the the wavelength of a photon or or whatever.

It's I mean, it is it is an amazing thing.

And It is.

It is.

It's extremely interesting also because, again, this is something that you can do only with two d materials, at least in an easy way.

Exactly.

Yeah.

And, this makes, this is really unique to the two d material world that other systems struggle to do.

Well, that's great, Antonio.

Thanks so much for coming on to the podcast and talking about, two d materials.

And, I hope that your your research, using machine learning and, all that stuff you've got in the lab goes well.

Thank you.

Thank you so much.

It's been a pleasure being here.

Discover cutting edge science in minutes.

IOP Publishing's new Progress In series, Research Highlights website offers quick, accessible summaries of top papers from leading journals, like reports on progress in physics and progress in energy.

Whether you're short on time or just want the essentials, these highlights help you expand your knowledge on leading topics fast.

Perfect for busy researchers, curious minds, and anyone who wants to stay informed without the deep dive.

To start reading, just type progress in series research highlights into your favorite search engine or follow the link in the notes for this podcast.

I'm afraid that's all the time we have for this week's podcast.

Thanks to Antonio Rossi for his fascinating insights into two d materials.

And a special thanks to our producer, Fred Ailes.

We'll be back again next week when Physics World's Margaret Harris presents the first in a series of episodes recorded at the Heidelberg Laureate Forum, a meeting that attracts some of the best and brightest mathematicians and computer scientists.

See you then.