Quantum tests show 1,000x efficiency gain for chemical computing in South Korea

Qunova Computing, a Daejeon, South Korea-based quantum computing firm, is the first to have achieved ‘chemical accuracy’ on a commercial quantum computing device using its platform-agnostic algorithm.

‘Chemical accuracy’ is the threshold of 1.6 millihartrees below, which results from quantum computing platforms that must remain considered solutions for real-world chemistry applications.

Quantum computing, the next frontier of computing technology, promises to revolutionize computational speeds and help us solve complex problems, from designing new drugs to optimizing logistical operations.

Quantum computing technology uses quantum bits or qubits to make calculations, and companies are building computers with a limited number of qubits to test the computations’ performance.

Errors compound easily in quantum computers built to date, with experts calling this the Noisy Intermediate Scale Quantum (NISQ) era. Yet, Qunova’s algorithm, dubbed HiQVE, aims to make these quantum computers deployable for real-world applications by improving their accuracy.

“We start to see several industrial users who are looking into running Qunova’s algorithm on NISQ machines to re-evaluate their previous works and trying to apply NISQ machine based calculations for larger problems,” said Kevin Rhee, founder and CEO of Qunova Computing in a conversation with Interesting Engineering ( IE ).

Achieving ‘Chemical Accuracy’

At the Quantum Korea 2024 Event, Qunova demonstrated chemical accuracy on a series of quantum computers, irrespective of their platform technology.

In the first demonstration, Qunova used an IQM 20-qubit machine and demonstrated its algorithm by producing energy estimations of three different geometries of Lithium Sulfide (Li2S) live at the event.

Before this, Qunova achieved an accuracy of 0.1 millihartrees using IBM’s Quantum Eagle processor, which uses 24 qubits. The company also achieved similar accuracy using the IBEX Q1 quantum computer, which uses 20 qubits and is built by Alpine Quantum Technologies (AQT), based in the European Union.

IQM and IBM use transmon superconducting architecture, while AQT uses trapped ions for its quantum computing platform. This also demonstrates that Qunova’s algorithm is platform agnostic.

“Quantum algorithms are platform agnostic if the algorithm performance is subject to quantum noise which can be unique for certain platforms. HiVQE is designed not to use the information associated with quantum noise,” explained Rhee in the email.

How does Qunova’s algorithm work?

Simulations used on conventional computers use Variational Quantum Eigensolvers (VQEs) to carry out complex computations. However, this approach could be more scalable and achieve chemical accuracy on quantum computers.

Qunova’s innovation lies in using a simplified version of VQE, which it refers to as Handover Iteration VQE or HiQVE.

“An analogy to explain this is like our algorithm takes information of what happened only, ignoring how many times it happened,” added Rhee in the email to IE. “The quantum noise often alters the information of how many times it happened, so Qunova is using only information that is unaltered.”

Its approach focused on not carrying over errors from quantum computations and excluded “Pauli word measurements,” usually used to measure the spin of qubits along different axes.

This allowed the team only to extract essential data related to the orbitals of each qubit, which were then fed into a classic computer to calculate the result quickly and with a higher level of accuracy.

Overall, the computations also ran 1,000x more efficiently. “This is a very exciting result for our team, and indeed for the quantum computing community more broadly,” said Rhee in a press release .

“These results show that we can meet the requirements of industrial users on existing NISQ machines. We anticipate that running a similar demonstration on a NISQ machine with as few as 40-qubits could provide industrial users with a real quantum advantage.”

The team will spend the next few months demonstrating this.