WHY THIS MATTERS IN BRIEF
Everything in our world is based on chemistry and chemicals, and quantum computers that can create new compounds and materials in seconds will revolutionise everything.
Love the Exponential Future? Join our XPotential Community, future proof yourself with courses from XPotential University, connect, watch a keynote, read our codexes, or browse my blog.
Using Google’s Sycamore quantum processor, scientists have performed the largest chemistry simulations on a quantum computer to date, and to do it they used a new technique that may help resist the noise often seen in quantum circuits.
A little while ago Google’s quantum computer achieved quantum supremacy and managed to perform a calculation that would have taken the world’s largest supercomputer Summit – which can do 200 Quadrillion calculations per second – billions of years to solve so giving these futuristic monster computers chemistry problems to solve that today even the best supercomputers find difficult to solve could revolutionise everything about chemistry, and then in urn revolutionise every single thing that uses chemicals, from fertilisers to materials, and beyond.
Furthermore, the more qubits that a quantum computer has, the greater its computational power can grow, in an exponential fashion, something that’s known as Roses’ Law.
The best near term application for quantum computers may be chemistry, such as simulating molecular reactions that might yield insights into next-generation batteries or new drugs. Performing these kinds of simulations becomes exponentially more complex as molecules get larger, which can prove an overwhelming challenge for conventional computers but might be something quantum computers can overcome.
In the new study, the group of researchers from Google Quantum AI, Columbia University, and the University of California, Berkeley, used a Monte Carlo algorithm, which essentially treats problems as games where solutions are reached via many random simulations of these games. Specifically, they relied on a Monte Carlo algorithm designed for quantum physics models of fermions, a class of particles that includes electrons.
Normally, fermionic quantum Monte Carlo algorithms run on classical computers do not scale well to simulating larger molecules. The researchers found that a hybrid approach that combined both classical and quantum computation could help their fermionic quantum Monte Carlo algorithm overcome this obstacle.
In experiments, scientists employed up to 16 qubits on Google’s 53-qubit quantum computer to calculate the ground state of molecules, the one in which they have the least amount of energy. A molecule’s ground state is influenced by factors such as the number of electrons it possesses and the paths these electrons take as they orbit a nucleus.
The researchers simulated the molecules H4, molecular nitrogen, and solid diamond. These involved as many as 120 orbitals, the patterns of electron density formed in atoms or molecules by one or more electrons. These are the largest chemistry simulations performed to date with the help of quantum computers.
A classical computer actually handles most of this fermionic quantum Monte Carlo simulation. The quantum computer steps in during the last, most computationally complex step calculating the differences between the estimates of the ground state made by the quantum computer and the classical computer.
The prior record for chemical simulations with quantum computing employed 12 qubits and a kind of hybrid algorithm known as a Variational Quantum Eigensolver (VQE). However, VQEs possess a number of limitations compared with this new hybrid approach. For example, when one wants a very precise answer from a VQE, even a small amount of noise in the quantum circuitry “can cause enough of an error in our estimate of the energy or other properties that’s too large,” says study coauthor William Huggins, a quantum physicist at Google Quantum AI in Mountain View, Calif.
In addition, a VQE “can also take a long time to perform enough measurements to get that very precise answer,” says study coauthor Joonho Lee, a quantum physicist at Columbia University, in New York City. “On top of this, we often have to optimize the parameters of our quantum circuit to prepare a good approximation to the ground state, and this can add an even larger overhead to the whole process.”
One potential concern with this approach is that qubits are fragile and prone to error. However, whereas VQEs require very little noise in quantum circuits in order to get very precise estimates of ground states, this new technique does not, which means “we can sometimes get away with more noise,” Huggins says.
“We’ve already blown past the largest VQE that folks have ever managed to perform, and we think that we can go significantly larger, even on the noisy quantum computers we have today.”
“In fact, we provided evidence in the paper that, even for our largest experiments, the noise on the chip wasn’t the limiting factor,” Lee says. “Rather, we weren’t ambitious enough with the design of the circuits that approximated the ground state. This tells us that we have a good shot of scaling up our current approach further even without developing new theoretical tools, and that’s a real beacon of hope given how hard it can be to perform accurate calculations of quantum chemistry on a noisy device.”
The new technique achieved a level of accuracy nearly as good as the best current classical method. In the future, they hope “we’ll make enough progress that attacking problems that pose a challenge for classical algorithms becomes practical,” Huggins says. Still, “at the end of the day, we expect it to be extremely challenging to get a practical advantage for quantum chemistry using the noisy quantum computers that we have today, or even tomorrow.”
The next step for the researchers is going to larger experiments “where we try to get to the limits imposed by the noise on the current generation of Sycamore,” Lee says. “As we make progress on developing and understanding new algorithms, we’re also expecting that new advancements in the hardware, and the software that controls it, will keep making our job easier.”
The scientists detailed their findings online 16 March in the journal Nature.
