IBM Brings Quantum Computing to Real Chemistry: Created and Simulated the 'Half-Moebius' Molecule
A team of researchers from British, Swiss, and German universities, along with researchers from IBM Research Europe, has published an article on a new molecule whose existence had previously only been hypothesized. But this is not the only groundbreaking achievement: to do this, the researchers relied on an IBM quantum computer, demonstrating how it is already possible to use these devices to conduct science with real and tangible discoveries. We spoke with Ivano Tavernelli, one of the IBM researchers who contributed to the study.
A Unique Molecule
The research, published in Science and available on arXiv, focused on the C13Cl2 molecule, characterized by a ring shape. Unlike benzene, however, it was not necessary to dream of the uroboros to understand its shape: the molecule was in fact artificially built using an atomic force microscope by manipulating individual atoms. This molecule has a property that makes it 'half-Moebius', meaning it is a 'half Moebius' molecule. The Moebius strip is a surface that can be created by taking a long strip of paper, flipping one end, and joining it to the other. What you get is a surface where one twist takes the outside to the inside and vice versa, necessitating two complete turns to return to the starting point.
A 'half-Moebius' molecule has a topological configuration of electrons with a similar property, but it will require four turns to again return to the starting point instead of two. In other words, each turn will result in a 90-degree rotation. To visualize the concept, one can imagine taking a square-section rubber band and twisting it around a finger: following this 'half Moebius' rubber band around the finger will show that it rotates 90 degrees with each turn, returning to the starting point (and the original 'face') after four turns.
The Importance of This Discovery
"It's fundamentally important: it concerns the experiment done with these instruments, so with an atomic force microscope. The existence of these molecules had been hypothesized, but they had never been synthesized, and this is what the group has managed to do for the first time: realize in practice a molecule that existed only theoretically," Tavernelli tells us.
"Whether these molecules have applications is harder to say, because they are synthesized under very particular conditions and at very low temperatures. However, one cannot exclude that one day they may be stabilized and then applied, for example, in quantum sensing, but there is no evidence in that direction yet."
The point of the study is that it is one of those cases in which something has been realized that did not exist before and was thought to be only a theoretical construct.
"It is one of those studies that can change textbooks," Tavernelli states.
Quantum Computers as a Simulation of Nature
But the purely chemical domain is only one side of the coin. The other consists of the use of quantum computers. Richard Feynman's quote is well-known: during a lecture, he stated that "nature is not classical: if you want to create a simulation of nature, you must do it quantum mechanically." And that is precisely what the researchers did: they used a quantum computer to simulate the characteristics of the 'half-Moebius' molecule.
"For the first time, we have demonstrated that it is possible to design and characterize these molecules using a quantum computer instead of a classical one. This means it is possible to do good science, such as discovering a new molecule and publishing it in Science, using a quantum computer," Tavernelli says.
On the left, the electron cloud density measured with the atomic force microscope; on the right, the simulation of the same carried out using a quantum computer.
Speaking of textbooks, this is likely one of those moments that will be recorded in them: for the first time, it has been possible to use a quantum computer to make a discovery in the chemical field. However, attention must be paid to one aspect: we are not yet at the point where quantum computers have done something that classical ones could not achieve.
"It simply means that a quantum computer can be used to do as well as a classical approach," Tavernelli explains. "We could reach the so-called 'quantum advantage'—conducting calculations that are beyond classical computers—by improving the efficiency of the quantum algorithm and applying it to even more complex molecules."
For now, the milestone achieved is nonetheless significant: as Tavernelli underscores, "the value [of the study] is precisely that we have reached enough accuracy to conduct research. At this level, it is comparable to other classical methods; it is superior to many approximate classical methods, and we have reached the level of the most sophisticated ones."
Another noteworthy aspect is more technical and technological. This study is a practical demonstration of the model IBM has been discussing for a few years, namely 'quantum-centric supercomputing': a synergy between quantum computers and classical supercomputers, where each performs calculations in which it offers the best performance. In this case, the results from the quantum computer were transformed into classical matrices, which were managed via a supercomputer.
The calculation would have been feasible, in principle, even by a classical supercomputer, but this would have required exploring an enormous combinatorial space (that is, a very high number of possible matrices) or using approximations. The role of the quantum device was instead to "select" the correct matrices (using technical jargon, by sampling the relevant subspace and constructing the representation of the system's Hamiltonian), thus reducing computation time while maintaining a high level of reliability without introducing approximations.
Looking to the Future: Between Quantum and AI
The field where AI has perhaps demonstrated its superiority so far and provided the best results is that of code writing: language models have proven to be very good at programming. This has led us to wonder: looking ahead a few years, what will be the impact of AI on programming for quantum computers? How much can AI help researchers focus not so much on quantum programming aspects but on those specifically related to their field?
"There are now many tools that can also produce new software, often of very good quality, based on what is already available (let's not forget that [AIs] do not invent anything new; they piece together existing elements). Will the same apply to quantum? I think so," Tavernelli tells us.
"There is definitely room for new algorithms even in this space. These tools will likely handle the coding part, hopefully more efficiently than we do, thus further accelerating the 'discovery loop', that cycle where a hypothesis is formulated, tested (with simulations or experiments), and updated based on results until an accurate predictive model of nature is reached.
There is no certainty about tomorrow, but what remains is that an important and interesting step forward has been made toward that future, talked about for years, where quantum computers will become protagonists in scientific research. This study demonstrates that this future is achievable and, in fact, is already here.