December 5, 2024 Pat Brans
Like other large German industrials, BASF is striving to achieve mastery of quantum computing as an end user—and it’s relying on partners, such as software startup Kipu Quantum, to help it get there.
BASF has been using high-performance computing to simulate chemical reactions on a molecular level for years. In 2017, it even bought its own supercomputer, Quriosity. “Our core competence is in chemistry and materials where we need to simulate molecules and chemical reactions to design new materials,” said Michael Kühn, principal scientist for next-generation computing at BASF.
While BASF raised the stakes by upgrading Quriosity to a second generation in 2023, chemists and physicists at the company have always known that classical supercomputers will never be able to deliver more than approximations when simulating physics. This has been known at least as far back as May 1981, when Richard Feynman first introduced the idea of building a computer on quantum mechanics in a talk called “Simulating physics with computers.”
Feynman pointed out that because nature is quantum-mechanical, when you try to simulate physics with numerical algorithms that solve differential equations on a digital computer, the best you can hope for is an approximation. Furthermore, as the size of the physical system grows, the size of the digital computer needed to approximate it grows exponentially. Feynman went on to say that “with a suitable class of quantum machines, you could imitate any quantum system, including the physical world.”
BASF has put a new supercomputer into operation at its site in Ludwigshafen. Like its predecessor, it is called Quriosity and is the world’s largest supercomputer used in industrial chemical research. With computing power of 3 petaflops, it enables significantly more complex modeling, virtual experiments and simulations. Quriosity has more than 1,000 computing nodes and 3,000 terabytes of storage capacity. Source: BASF SE.
How quantum computers might simulate chemical reactions
In 2017, BASF began to explore quantum computing. In 2021, the company helped found the QUTAC Consortium in Germany and has been collaborating with 12 other large German industrials that are experimenting with using quantum technology to solve real-world problems. Quantum computing could accelerate research and innovation by allowing BASF to make better predictions about new materials and chemicals. It would give the company a better understanding of what’s happening on a microscopic scale, leading to better material designs while minimizing the need for expensive lab trials.
Before a molecule can be simulated on a quantum computer, it must be specified on a classical computer—including the atoms that make up the molecule and their configuration. “Then you generate a starting state on the classical computer, which is mapped from the classical computer to the quantum computer,” Kühn explained.
A molecule, of course, consists of electrons and atomic nuclei—and the electrons live in molecular orbitals. All of these things are mapped to qubits. With the right quantum algorithm, the qubits mimic the orbitals and their interactions and thus the molecule. “There’s a correspondence between the molecule, which is a quantum mechanical object, and the quantum computer, which also relies on quantum mechanics,” Kühn said. “This is why it’s much more efficient to simulate a molecule on a quantum computer, as compared with simulating it on a classic computer.
“Once you have quantum computers that are error-corrected, which is currently not the case, you would then get the exact solution—for example, the exact energy of the molecule, if that’s what you’re solving for,” Kühn added. “If you want to simulate a chemical reaction, you repeat this calculation for each of the molecules involved, and by simple subtraction and addition, you could calculate the energy consumed or released by that chemical reaction.”
Part of BASF’s exploration involves finding suitable algorithms, an area where it gets help from partners such as Kipu Quantum, a startup that develops software tailored to a specific computer to solve a given problem with lower overall error rates. Typical quantum algorithms require a large circuit depth, a situation wherein a lot of gate operations are carried out in sequence, leading to a significant accumulation of errors. By reducing the number of gate operations needed for a given algorithm—or by having more gate operations performed in parallel—an algorithm not only executes more quickly but does so with fewer errors.
Kipu has found ways of minimizing circuit depth so a given problem can be solved on current hardware with a higher degree of accuracy. “We reduce algorithms by a factor of 100 to 10,000, depending on the application,” said Daniel Volz, co-founder and CEO of Kipu Quantum. “That essentially means we are much less affected by noise buildup.”
The power of end users in the global ecosystem
Volz said that German companies are very well-positioned on the end-user side, with virtually all of the big industrial players experimenting with quantum computing. “Compare that with the level of engagement in the United States,” he said. “The likelihood of a company being in quantum is higher in Germany than in any other place on the planet.”
With all the big end users embracing quantum technology, the demand side is full speed ahead, but Germany comes up a little short on the supply side. “It’s a shame, because a lot of public money went into fostering the quantum ecosystem—including the Munich Quantum Valley—yet there’s not a single very large, very advanced quantum company to show for that,” Volz said. “There’s a very strong academic ecosystem, but so far, there’s no significant player in quantum computing hardware.”
Nevertheless, Germany may end up with a very strong hand, because applications may be where the real payoffs lie. Currently, BASF has no ambitions to build its own quantum computer, but the company is intent on learning how to use them—and on mastering the algorithms that apply to its business. While the quality of available hardware was lacking in the past, that may be changing. “Since last year, we’ve seen remarkable progress in the field of error correction, which brings us closer to fault-tolerant machines,” said Horst Weiss, vice president of next-generation computing at BASF.
Weiss thinks companies such as QuEra and IBM could deliver fault-tolerant quantum systems by the end of the decade, but now he’s concerned about the state of the software. “So far, we haven’t seen an algorithm that would outperform our traditional machine for a quantum chemical problem,” he said. “We need smarter algorithms that can at least compete with classical computers for real applications.”
For Volz, Kipu is on the right track to delivering those algorithms. If things go as he predicts, the quantum advantage will be achieved first by a software company (namely his)—and that might occur within a year. He cautioned that it will be a narrow quantum advantage, where for a specific problem in a specific industry, a specific quantum algorithm and computer perform better than a classical machine. “There has already been enough hardware progress,” he said. “Now, it’s about bringing the hardware and algorithms together to make it over the tipping point.”
Over time, this narrow quantum advantage will lead to a broader quantum advantage, where a larger set of problems are better solved by quantum computers and algorithms. When that occurs, Volz expects an era of scarcity to ensue, characterized by two bottlenecks. One will be a shortage of talent: Every large corporation will start hiring several quantum physicists, sweeping the market clean very quickly. The level of sophistication of solutions will be such that a team of skilled data engineers will not be able to do the job themselves—without quantum physicists, a company will not be able to take advantage of quantum computing.
The second choke point will be access to hardware. But Volz said Kipu Quantum’s software will minimize the effects of hardware scarcity because it’s compact and requires less computing time. “Our algorithms require 500× less time on quantum computers than other algorithms,” he said.
Comments