How to use sample-based quantum diagonalization on IBM hardware

Molecular simulation is a crucial tool for scientists. By understanding the atomic-level behaviors of molecules and materials, we can uncover new drugs, catalysts, or other chemicals. However, trying to simulate the electronic structure of a molecule can tax even the most sophisticated classical computers.

Thankfully, as physicist Richard Feynman predicted, quantum computers have the potential to emerge as powerful tools for modeling the quantum structures found in nature. And IBM Research scientists have found that large-scale molecular simulations are possible with quantum computers running an algorithm called sample-based quantum diagonalization (SQD).

In the latest demonstration of the power and utility of SQD, researchers from IBM, Cleveland Clinic, and Riken used a variant of the technique to accurately model complex proteins, including one with more than 12,000 atoms. This breakthrough was announced this week at IBM Think, just four months after researchers achieved a prior milestone where they modeled a 303-atom protein.

Check out this explainer video from IBM Research to learn all about how and why SQD is such a powerful tool for scientists.

Electrons in atoms and molecules occupy orbitals associated with different energy levels. Orbitals are mathematical functions that describe the spaces around a nucleus where electrons are most likely to be found. Calculating the ground state of electrons in a molecule — their most stable, lowest-energy configuration — can yield insights into the molecule’s expected behavior and reactivity. It sounds simple enough, but electrons exert a physical influence on one another. All of these interactions mean that as the number of electrons rises, calculating the ground state becomes exponentially more complex, far beyond the capabilities of classical high-performance computers.

That’s where SQD comes in. Using quantum-centric supercomputing (QCSC), scientists can deploy this method with classical hardware in harmony with quantum processors to estimate the ground state properties.

To begin, a classically-computed structure of the molecule in question is mapped onto a quantum circuit, which is prepared to run on the specific quantum processor being used.

The quantum hardware then executes the circuit, generating a suggested set of configurations to explore with classical hardware. Classical hardware then performs a diagonalization operation on the set of configurations to produce an approximation of the system’s ground state. Some properties of the approximate ground state can be used to further improve the quality of the quantum hardware outcomes.