Or as Richard Feynman, the physicist regarded as one of quantum computing’s early visionaries, once put it: “Nature is not classical, damnit, and if you want to make a simulation of nature, you had better make it quantum mechanical.”
Scientists have long grappled with how to efficiently predict molecular reactions. Chemists today often use mass spectrometry – an analytical tool that helps measure the mass and charge of ion molecules – in combination with computational methods but the process can be cumbersome and inconsistent.
In addition to drug development, research labs strive to accelerate chemical reactions to uncover more efficient means of making energy and fuel, pulling apart atoms to determine whether a stimulating agent could be added.
This is known as catalysis, and automating it is one of quantum computing’s biggest holy grails.
Because quantum computing bits in essence represent the same individual pieces of matter – such as atoms, ions, electrons or photons – that comprise our universe, they should, at least theoretically, store information in the same way as chemicals do, and this could potentially strip away the scientific guesswork to deliver new catalysts faster.
Ammonia is a good example of the commercial value that could be realised. Since the invention of the process used to synthesise ammonia – the Haber-Bosch – in 1913, it has served as the precursor to agricultural fertilisers, greatly improving farm yields and therefore global food availability.
Above: IBM’s seven-qubit device from 2017
An estimated 85% of the global ammonia supply goes to fertiliser, helping to feed roughly half of the world’s population, according to a paper in the January 2018 edition of academic journal Johnson Matthey Technology Review.
Producers synthesise ammonia by exposing nitrogen from the atmosphere to hydrogen molecules to spark a chemical reaction, but the process uses plenty of energy and was responsible for pumping out some 451 million tonnes of carbon dioxide in 2010, according to chemistry news publisher C&EN, equating to around 1% of global emissions.
Building a quantum computer that intrinsically imitates ammonia’s atomic structure could speed up efforts to find catalysts considerably. Scientists are already trying – one paper in the Journal of Physical Chemistry claims nickel-doped iron was effective, based on a quantum mechanics-based screening experiment.
James Wootton, a research staff member at computing technology group IBM’s Research lab in Zurich, said chemistry and optimisation in finance or supply chains are likely to be ahead of sectors such as computer security in terms of commercialisation.
He predicted: “In areas like quantum chemistry – you are trying to calculate an energy with a value, and you have to find that value for it to be correct.
“More concrete use-cases might come a little bit later on than today, but I think as these applications are closer to the physics of the end-device, things like quantum chemistry will probably be before some of the more abstract ideas, like for example cracking RSA-based [Rivest-Shamir-Adleman] data encryption.”
BCG, the US-headquartered consulting firm, has pegged the value to end users from quantum chemistry simulators at around $500m from 2018 to 2022. Clients can expect to benefit in the near-term by paring research and development costs, thanks to efficiencies in designing new chemicals.
Later, whole new materials unearthed using broad quantum advantage – techniques wholly unattainable with classical machines – could fundamentally change the paradigm.
A revolution in energy generation, construction and transport may well follow, with a January 2020 BCG article arguing that quantum computing-derived technologies could one day address the source of most global carbon emissions.
Markus Solibieda, managing director of BASF Venture Capital, the corporate venturing arm of chemical producer BASF, says the firm has been tracking quantum computing for some time having made its investment debut in early 2019 through Harvard University-founded quantum software developer Zapata Computing.
He said: “With our internal experts, we have watched the space for many years. We are looking at developments in materials science and at how new computational power or concepts can help accelerate our research and development.
“Chemical research requires a lot of computational power because you are dealing with multivariable optimisation problems.”
“Quantum computing has always been an area of interest for BASF Venture Capital, but only in the last two years have we reached the conclusion that it will be something that is here for real, and that will materialise faster than expected.”
As mentioned in GCV’s last quantum report, Google published results of a basic chemistry experiment performed on a quantum computer in August 2020 to simulate reactions in a molecule called diazene.
Google’s team started by simulating a basic version of diazene’s energy state, using a machine with 12 quantum bits, or qubits – the unit of quantum information that equates to the binary on-and-off switches, or bits, used to operate classical machines.
Unlike bits, qubits are able to employ a quantum-mechanical phenomenon called superposition to store more complex inputs, enabling more versatile calculations. In Google’s experiment, each qubit was calibrated to represent a single electron inside 12 hydrogen atoms, according to Scientific American.
The model was then told to emulate a chemical reaction between hydrogen and nitrogen atoms – the structure of diazene contains both – to predict how the electronic structure would adapt.
Google’s achievement builds on a six-qubit chemistry simulation undertaken by IBM in 2017. In an interview with Scientific American, Ryan Babbush, the research lead for the project, suggested the experiment had matched the performance of an early computer in the 1940s.
He added: “If we double it again, we will probably go to something like 1980. And if we double it again, then we will probably be beyond what you could do classically today.”
Above: BASF’s supercomputer Quriosity
To achieve true quantum advantage – solving problems that existing classical machines cannot resolve – Google’s feat needs to be expanded to reliably process larger and more lucrative compounds.
Solibieda said he expected this to come in a matter of years as the number of qubits incorporate machines is expected to rapidly increase, adding that 100 qubits is generally seen at the point at which commercial-grade services could be sustained.
Solibieda explained: “While two years ago, there were experiments on the hardware side where you would talk about three or four or half a dozen qubits, now you are talking about systems in R&D that are up to 50-80 qubits.
“Most experts would say that once you reach 100 qubits or more then this would be the level when you can actually commercialise quantum computers and find broader applications to use these major advantages in terms of speed and computational power.”
The big question is which architecture will achieve the stability needed to consistently execute chemistry calculations at industry standards.
Google exploited ultra-cool superconducting wires that offer zero electric resistance to conduct its chemistry experiment, an approach also favoured by IBM for much of its internal R&D program.
Researchers at University of Science and Technology of China recently claimed to have achieved quantum advantage by manipulating photons – individual particles of light – although reports have suggested the technology cannot be reprogrammed for different tasks.
In addition, caution must be applied to any statement of quantum advantage. Google has claimed the same watershed for a basic calculation before, only for IBM researchers to pour cold water over its achievement by apparently executing the same experiment on classical technology.
Unlike photon-powered computers, spin qubit-powered machines use electrons trapped in what Wootton describes as “exotic” semiconductors to store information.
One might expect them to be hypothetically cheaper to operate as a result, however a key pitfall with spin qubits is that their natural interactions do not align precisely with quantum algorithmic instructions, making programming more challenging, said Wootton.
Chemists might find this sufficient so long as the experiment is justified by a relatively profitable use-case, but Wootton argues superconducting circuits are proven for more applications. Ultimately, it really is too early to say which architecture will emerge as the winner.
Wootton explained: “The number of qubits you need to run a fault-tolerant algorithm is in the millions.
“So I think in that regard superconducting circuits could be seen as relatively large, whereas things like spin qubits which we are also looking at, tend to be smaller.
“It may be that you want to switch something that is something space-efficient, but superconducting circuits really seem to already have a lot of pros to them.”
At BASF’s existing supercomputer in Germany, capacity is fully booked owing to a stream of requests for complex chemical investigations.
BASF has leveraged this expertise to assist with the practical side of quantum compute, helping engineers ensure architectures and algorithms meet commercial needs right from launch.
Solibieda said: “We have the knowledge of the exact kind of mathematical challenge and optimisation problems, for example in the research for inventing a new material.
“In the past there has been very much trial and error in chemical research, and you want to reduce this by narrowing down the options.
“So now we are communicating to the quantum computing world, to the experts and the scientists, what we most urgently need in the chemical industry.
“These experts and scientists then tell us what a quantum computer algorithm can do today, and we convey what we need in terms of optimisation.
“We try to find common ground, to expand the area so we can say it is not just this one problem where the quantum computer is the best solution, but to ultimately increase the scope of applications for quantum computing.”
At the other side of the equation, IBM has been working out how to support clients including oil producer Exxon and air carrier Delta Airlines to solve physics problems on its quantum computers using its cloud-hosted service, as part of its IBM Q Network programme.
A big question it faces is identifying the best mode for enabling the end-user to run applications on its quantum systems.
Standard computers can be operated through logic gates – circuits hard-wired to perform fundamental coding functions –
but controlling quantum machines is more nebulous as instructions have to be translated into electromagnetic pulses before qubits can understand.
A pertinent question, and one worth of discussion in GCV’s next special quantum computing report, is what level of quantum physics knowledge will be needed to program these game-changing machines.
Wootton concluded: “We are looking at different ways that end-users can deploy the quantum equipment – for example there are people who use it at the level of physics, but also clients that want to just feed the computer a problem using standard programming languages.
“So the question seems to be either hiding away all of the quantum machine level gates so that people do not even have to look at it, or allowing that kind of access.
“We have kind of high-level programming languages in classical computing which lie between machine level and a compiled algorithm, but quantum computing is more modular – with multiple different components that can be used to create a quantum advantage.
“Making an algorithm is often a case of putting these together – so while I do not think you will have to have a new programming language, you will have to know which functions are available to you.”