The Quantum Theorist at the Wedding
At the weekend I went to a wedding and sat next to a theoretical physicist. This was exciting enough, but I discovered that he'd just finished a PhD in an area of research very similar to my own (computational quantum mechanics). It’s a pretty niche area so I was astonished another human being had actually heard of it. Naturally I wanted to talk to find out more.
As he explained his work I realised the discovery he’d made was truly phenomenal. A potential game-changer which alters our understanding of Physics. Then, as I finished worshipping him, the conversation went like this…
Him: So it turns out the spin-state energies for non-ground-state systems follow a Gaussian distribution and that small-to-large system interactions lead to maximum entanglement.
Me: That’s amazing!
Him: Thank you, what was your research on?
Me: I made liquid wood.
Me: So, how’s your cheese-and-flower soup?
His work was so technically advanced I can’t sum it up neatly in a weblog, other than to say: we thought particles did a thing, but it turns out they do a different thing! This guy was making fundamental discoveries on the forefront of modern quantum mechanics, but my research can be summed up in a single sentence over a wedding lunch...so who’s the real genius here?
It then occurred to me that I’ve never actually talked about my research. I mention it briefly on the “about me” page and I’ve occasionally discussed it in lessons because I honestly think it’s interesting (that’s why I studied it) but I’ve never gone into any detail. So here it is for them what is curious.
Why would you want to make liquid wood?
As a species we use about 2x10^20 Joules’ worth of energy per year. Most of that comes from burning coal, oil and natural gas. Three main problems with this. 1) burning these substances is toxic, 2) they fill the air with greenhouse gases, 3) they will run out. We need a new energy source and we need it badly.
A potential candidate is cellulose: one of the main substances found in plants. For starters, when you burn cellulose it burns clean i.e. doesn’t produce any toxic chemicals, second you can replant the trees you cut down (making it carbon neutral) and third, plants will be around for a very long time. Cellulose solves all three problems and the good news is we have lots available. The amount of cellulose energy produced every year by nature is around 2x10^21 i.e. if we harvest just 10% of what nature already produces, we solve the energy crisis. Nice.
Unfortunately we can’t just burn wood as a civilization (for a million reasons) so we need to extract the cellulose from the plant-matter and burn it on its own. Just one problem: Cellulose is VERY hard to extract from plant. It’s like mixing up a bunch of candyfloss with a bunch of carpet-fibres and then trying to extract the candyfloss.
What we really need is to somehow turn wood into a liquid (liquids are usually easy to separate).
Step One – What’s already been done?
If you want to do good research you have to find out what other people have tried. And, it turns out, somebody had already started investigating liquid cellulose. In 1934 a man named Charles Graenacher discovered, purely by accident, that you could put small amounts of wood into a novelty chemical called an ionic liquid and get a thick wood-like sludge at the end. Back then his discovery was considered a bit of a joke – what would you want liquid wood for? It’s only in the past decade that people have rediscovered Graenacher’s research and realised how potentially useful it could be.
Thing is, we discovered the process by accident so it was basically a matter of fumbling around in the dark. Nobody knew why wood dissolved in these ionic liquids. If we could work out what was going on however we might be able to come up with a better way of doing it. After all, you can’t launch a rocket until you understand how gravity works.
There was already a bit of work being done on wood solvents by the time I got interested in the topic, but no theory behind it. Essentially it was a matter of: chuck some wood into a chemical and see if it works. Luck in other words. And this is where I began.
Step Two – Invent some calculations
Molecules are insanely small. Too small to be seen with a microscope. We also can’t ask them “how are you guys interacting?” The only way to know how molecules are really behaving toward each other is to use something called computational quantum mechanics.
Quantum mechanics is the most fundamental theory we have for explaining the world. It’s the study of the Universe at its deepest level – the very core of understanding. The only problem is that quantum mechanics is less than a hundred years old and, although rigorously tested and validated, we don’t know everything about it.
So, firstly I had to invent a new quantum-mechanical method for calculating how the molecules in wood interacted with the molecules in ionic liquids (if you want a more technical description of how I did this see below*).
This took several months and it was probably the most exciting part of the research. I spent most of my time sitting in front of a computer combining different equations to see if I could get any which worked. I can’t pretend I had a hunch from the beginning and I can’t pretend I had a Eureka moment either. Really, it was months of educated guesswork which finally started to yield sensible results.
So, about three months in, I developed a new way of simulating particle interactions by bootstrapping a bunch of different quantum mechanical methods together. Honestly, the method I invented is pretty clunky. It can be summed up in a single equation (let’s call it the tim James equation) but it looks so ridiculous I’d feel embarrassed to even display it. But, ugly as it was, my new equation worked to 96.6% accuracy.
Step Three – Use quantum mechanics to work out what’s going on
With my brand new method for calculating interactions between molecules I began teaching the computer how to correctly simulate the chemicals in wood and the chemicals which dissolved it. This would hypothetically tell me what was going on at the quantum level.
This part of the research was mostly tedious data gathering. I’m talking: wake-up, input some information, wait for the computer to pump out an answer, then repeat, over and over. For eternity.
My calculations also took up a lot of computer-space, so I decided to run everything at night. I became nocturnal because it was the best time to get access to all the computer-power I needed. Furthermore, to save time, I rigged nine computers to run simultaneously on different parts of the calculation. Many is the night you could find me in the computer lab at three in the morning, nine-computers all humming together as I ran back and forth between them making sure none of them went wrong. And I took a couple of photos...
After several months of this I began to start seeing a pattern. I got a feeling for what was going on with all these molecules...according to my computer simulation of them...and I started to make a guess as to why wood can be partly dissolved in certain ionic liquids.
Cellulose strands are tightly bound to each other via something called Hydrogen bonding (it's the same thing which makes the kevlar in bullet proof vests so tough). Certain ionic liquid chemicals seemed to be the right shape, size and charge to slip between the cellulose strands and ease them apart.
Step Four – Test it
My hypothesis suggested that for wood to dissolve, the ionic liquid had to have certain key features. So I decided to put them all together in one molecule. This new chemical I designed was named 1-crotyl-3-methylimmidazolium chloride (CMIM Chloride for short) which, if my hypothesis was correct, would out-dissolve any other ionic liquid.
And this is where things got tense. By now I’ve spent about a year on the project. I’ve invented my own equation, my own QM-approach, my own hypothesis and my own chemical. But if I don’t actually test it I might as well just say “I reckon this is true...” I had to actually make my chemical for real in the laboratory, sprinkle in some woodchips and see if it worked.
This is the part I suck at by the way. I’m not a very good lab chemist. Perhaps I’m just not very patient, or maybe too clumsy or careless, but I’m much safer with a computer and pen/paper in front of me than glassware. Everybody else is much safer too.
However, with a few false-starts, a lot of cursing and one near-explosion (sadly I’m not joking, I really did once almost blow up a fume cupboard by accidentally making a bomb) I finally managed to synthesise my computer-predicted wood solvent.
And then the moment of truth. It worked.
We tried using wood-chips, cellulose powder and even cigarette paper and they all sunk into the ionic liquid and formed a liquid. The new chemical really did dissolve wood about as accurately as my model predicted.
By no means does this prove my theory is correct of course, someone could come along tomorrow and point out the flaws in my method, but it was a really nice way to finish. Technically, even if my hypothesis had been proven wrong then that's still good for Science. Any discovery tells you something, even a negative one. But I'm a human...and it felt nice to have my hypothesis validated. And that's where my work ended.
I did take a photograph of it but I can't find it anywhere. If you're curious however, this is a photograph of liquid wood from another research group...
So there we have it. I managed to create liquid wood and not die. There are still some big problems with my chemical however. It cost about £60 to make 10 centimeters cubed of the stuff, as well as 5 hours of dangerous labwork. And then the actual dissolving of the wood took about an hour. My chemical is expensive, works slowly and takes ages to prepare (not to mention being quite dangerous) but making an efficient wood solvent wasn't really my aim. My aim was to come up with a theory which would help future Chemists design their own wood solvents.
Now that we know how it works (assuming I'm right...fingers crossed) we're not fumbling in the dark any more, we can go straight to the design stage. My contribution to this field is minor but I'm still proud of it. If, one day, we find a way to base our planet on a cellulose economy rather than a fossil fuel one, I can say that I helped, in a very small way, to nudge us in the right direction. Basically I've helped save the world. You're welcome.
* Technical bit...
Unfortunately you can’t fully describe what a particle is doing at any given time due to the Heisenberg uncertainty principle (it is impossible to simultaneously know the momentum and location of a sub-atomic particle). Instead, we have to work out the probability of a particle being in a certain place at a certain time. We can do that using the Schrodinger wave equation. This treat’s the particle’s probable behaviour as a wave. A wave which describes the probable behaviour of a particle is called a “wavefunction”. If you calculate the wavefunction for a particle you can predict its probable behaviour.
Currently, the mathematics required to accurately model anything more than two particles just doesn’t exist. We simply haven’t figured out how to solve the Schrodinger equation for a three-body system. The reason is that every time you add a particle in, you add a huge number of extra parts to the calcultion. For example, to correctly solve the Schrodinger equation to model a single atom of Iron requires (I have genuinely counted out the zeroes here, this isn’t just a joke) 10000000000000
00000000000000000000000000000000000000000000000000000000000000000 different terms in one equation. No computer on Earth can do this yet.
By the late 80s however, a new type of quantum mechanics was being developed called Density Functional Theory (DFT for short) which uses a few clever approximations and fudgey-guesses to give you an answer which is pretty reliable. So if it’s a choice between “perfect answer but impossible to calculate” and “reasonable answer and can actually calculate” the second option is the clear choice. Therefore, most people modelling many-body systems will be working with DFT.
DFT, invented by Walter Kohn, takes a completely different approach. Rather than calculating the wave-behaviour of each particle individually and combining them all, Kohn suggested we treat all the electrons as one thing: a sort of fuzzy electron-cloud which can be thick in some places and thin in others. This 3D cloud of electron-ness tells you the probability of finding an electron in a certain place…thicker the density the more likely it is to be there.
DFT is still pretty new though, and at the moment it can only calculate the electron density for one molecule, not an interaction between several. So here’s what I actually did. DFT works as follows: you start with an approximately correct structure (often based on an earlier, cruder type of sum called a Hartree-Fock calculation). You then undergo what’s called an iterative process where you make a little change to the molecule, calculate the stability, make another change, calculate stability, make another little change, calculate and so on and so on until you get an optimum answer which can’t be improved or made any more stable.
Sometimes you go in the wrong direction of course and head away from stability so your final answer ends up with a molecule whose stability is the square-root of minus one. (Not that I ever did that of course, never.)
What I decided to do was combine DFT with another type of calculation called molecular mechanics (MM). MM is very good at dealing with many-body systems which is DFT’s main weakness. So I spent a few months trying to find a way of combining these two approaches to create some kind of super MM-DFT process. To describe the actual process I used to develop this approach would be very tedious so I'll just go to my final method which worked as follows:
Start by assuming all the atoms are rubber balls and all the bonds between them are springs (treat them as classical structures in other words). Put them into a virtual 3D box and bounce them around using MM several thousand times. Eventually, by pure chance, it will end up finding stable and likely arrangements of all the molecules. Take this structure and perform hundreds of little calculations on it with DFT until you end up with a sensible answer. Simple as that. Nothing fancy, nothing groundbreaking, just take your answer to one calculation and use it as the starting point for the next.
The only problem still arising was that DFT treated the whole system as one molecule rather than three. To get rid of this problem I used an even cruder fudge-factor where I calculated the average difference between a covalent bond and an ionic bond and simply subtracted this from every bond which shouldn’t have been there. It’s honestly amazing how close to the real answer this ridiculous cheat got me. But I can honestly say I’m the first person in the world to have used it. I’ll probably also be the last.
Seriously, if equations are works of art, mine was a Jackson Pollock painting.
I love science, let me tell you why.