Caltech’s chemistry professor Sandeep Sharma, along with associates from IBM and the RIKEN Center for Computational Science in Japan, is offering a preview of the computing landscape ahead. The group has leveraged quantum computing alongside classical distributed computing to tackle a particularly intricate issue in quantum chemistry—ascertaining the electronic energy levels of a comparatively complex molecule. This endeavor showcases the potential of such a quantum–classical hybrid strategy for propelling forward not just quantum chemistry but also domains like materials science, nanotechnology, and drug development, where understanding the electronic characteristics of materials can reveal their expected behaviors.
“We have demonstrated that it is possible to integrate classical algorithms functioning on high-performance classical machines with quantum algorithms running on quantum devices to achieve valuable chemical insights,” remarks Sharma, a recent addition to the Caltech faculty whose research centers on creating algorithms for investigating quantum chemical systems. “We refer to this as quantum-centric supercomputing.”
Initially, the team employed an IBM quantum instrument, powered by a Heron quantum processor, to streamline the calculations, subsequently utilizing RIKEN’s Fugaku supercomputer to address the problem. They harnessed as many as 77 qubits, or quantum bits, during this process. Most preceding quantum computing attempts aimed at exploring chemical systems have managed to utilize only a limited number of qubits.
The study appears in the latest edition of the journal Science Advances and is highlighted on the cover.
In this latest investigation, the researchers aimed to utilize quantum computing to explore an iron–sulfur system, the [4Fe-4S] molecular cluster, which is a crucial constituent of various biological reactions. For instance, the iron–sulfur cluster seems to play a pivotal role in the enzyme nitrogenase, responsible for nitrogen fixation—the transformation of atmospheric nitrogen gas into ammonia, which enables plant growth.
Determining a System’s Wave Function
In quantum chemistry, a fundamental value that unlocks a wealth of information about a system is its ground state, the lowest achievable energy level. Should chemists wish to ascertain if a system like the [4Fe-4S] molecular cluster is reactive, evaluate its stability, or comprehend its role as a catalyst, they must first establish the ground state. In this state, a mathematical representation referred to as a wave function delineates the likelihood that the electrons within the system will occupy a specific location. Solving a quantum-mechanical equation known as Schrödinger’s equation yields this wave function.
Many classical algorithms face challenges in determining the accurate wave function of the [4Fe-4S] molecular cluster.
Over the past decade, it has become somewhat of a principle in the realm of quantum computing that resolving the wave function of such iron–sulfur molecular clusters via a quantum algorithm will illustrate that these algorithms can surpass their classical equivalents.
“The current document doesn’t completely establish that we surpass any classical algorithm, but it significantly advances beyond what has been achieved in the past with quantum algorithms,” Sharma notes.
Quantum Can Recognize the Essential Values in a Matrix
Researchers typically supply a classical algorithm with all known details about a system—such as the number of electrons, the atoms’ positions, etc. From this information, an algorithm generates an extensive matrix called the Hamiltonian. Its dimensions grow exponentially with each added electron in the system. “You can quickly find yourself facing computational limitations,” Sharma explains. However, many values within these matrices are not essential for calculating the system’s wave function. Consequently, classical algorithms often resort to a type of approximation termed classical heuristics to condense the matrix, creating a smaller, more digestible subset of representative values that are simpler to handle. Sometimes these approximations succeed, and sometimes they do not.
“This study illustrates that those classical heuristics can be supplanted by something more methodical that originates from a quantum computer,” Sharma clarifies. “Instead of relying on classical heuristics, a quantum computer is identifying the most significant components within the matrix.” Accordingly, the scientists input the most pertinent segment of the Hamiltonian matrix into the RIKEN supercomputer to compute the precise wave function.
The paper is titled “Chemistry beyond the scale of exact diagonalization on a quantum-centric supercomputer.” Besides Sharma, the lead author of the research paper is Javier Robledo-Moreno from IBM. Robledo-Moreno, along with IBM colleagues Mario Motta and Antonio Mezzacapo, serves as the corresponding authors. Additional contributing authors from IBM include Holger Haas, Ali Javadi-Abhari, Petar Jurcevic, William Kirby, Simon Martiel, Kunal Sharma, Iskandar Sitdikov, Kevin J. Sung, Maika Takita, and Minh C. Tran. Contributors from the RIKEN Center for Computational Science in Japan are Tomonori Shirakawa, Rong-Yang Sun, and Seiji Yunoki. The research received favorable support from the New Energy and Industrial Technology Development Organization and the RIKEN TRIP initiative.