Quantum Algorithms Now Model Complex Chemistry with Greater Accuracy

Ernst Dennis Lægteskov Binau Larsson and colleagues at University of Southern Denmark modelled complex surface chemistry using quantum computing, addressing a long-standing challenge for traditional methods. They benchmarked state-averaged factorized unitary coupled cluster with singles and doubles (SA-fUCCSD) and the adaptive, problem-tailored ansatz (SA-ADAPT) for modelling NO adsorption on Rh-doped TiO₂, a system exhibiting sharply multiconfigurational character. The findings reveal that adaptive ansätze offer improved efficiency and accuracy compared to conventional methods, requiring fewer computational resources to achieve results comparable to state-averaged complete active space self-consistent field (CASSCF) theory. This provides a key benchmark for evaluating quantum algorithms applied to realistic, chemically motivated systems and moves beyond simplified models, enabling more accurate simulations of catalytic processes.

Important operator reduction maintains accuracy in multiconfigurational quantum chemical

SA-ADAPT achieves near-CASSCF accuracy using 43% fewer operators than SA-fUCCSD, a reduction previously unattainable with conventional quantum ansätze. This advance enables accurate modelling of complex systems exhibiting strong multiconfigurational character and multiple state crossings, such as NO adsorption on Rh-doped TiO₂, where traditional methods struggle with localized electronic states. The inherent difficulty arises from the need to accurately describe the simultaneous interaction of multiple electronic configurations, demanding exponential scaling of computational resources with system size for traditional wavefunction methods. Density functional theory (DFT), while computationally cheaper, often fails to accurately capture strong correlation effects and charge transfer, particularly in systems with transition metal dopants like rhodium. An embedded cluster model retained SA-CASSCF orbitals to isolate wavefunction ansatz performance and benchmark both SA-fUCCSD and SA-ADAPT against this established method. The embedded cluster approach focuses computational effort on the region of chemical activity, in this case, the Rh dopant and the adsorbed NO molecule, while treating the surrounding TiO₂ lattice with a less computationally demanding method, reducing the overall system size without sacrificing accuracy in the critical region.

A modified operator selection scheme further accelerated convergence, indicating the potential of adaptive ansätze to efficiently tackle challenging multistate problems in quantum chemistry and beyond. The operator selection process is crucial for determining which electronic excitations are included in the wavefunction expansion. Traditional methods often employ a fixed set of operators, leading to inefficient use of computational resources. The modified scheme dynamically selects operators based on their contribution to the overall wavefunction, prioritising those that have the greatest impact on the energy and properties of the system. These results highlight substantial gains in computational efficiency, but currently rely on a fixed orbital basis and do not yet demonstrate the full potential of these ansätze when combined with adaptive orbital optimisation or applied to larger, more complex systems. Adaptive orbital optimisation would allow the orbitals themselves to change during the calculation, further improving the accuracy and efficiency of the method. The improvement in computational efficiency was observed when modelling NO adsorption on Rh-doped TiO₂, a system known for its complex electronic behaviour and multiple state crossings. The adsorption of NO on metal oxide surfaces is a fundamental step in many catalytic processes, including those used in automotive exhaust treatment and industrial ammonia synthesis. Understanding the electronic structure of this system is vital for designing more efficient catalysts. Incorporating multiple operators during each calculation step, the modified operator selection scheme accelerated convergence, and the embedded cluster model allowed for a direct comparison against the SA-CASSCF reference method. Convergence, in this context, refers to the point at which the calculated energy and properties of the system no longer change significantly with further computational effort.

Evaluating quantum algorithm performance for correlated catalytic systems

Accurately simulating the behaviour of electrons in complex systems is crucial for catalysis and materials science, yet current computational methods often fall short when dealing with materials exhibiting strong correlation and multiple possible electronic states. Wavefunction-based approaches offer a path beyond the limitations of density functional theory, but historically, the sheer computational cost has restricted their application to relatively small models. The computational cost of traditional wavefunction methods scales factorially with the number of electrons, making it impossible to treat even moderately sized systems with high accuracy. This demonstration of adaptive quantum ansätze represents a promising step forward, establishing a benchmark for evaluating quantum algorithms on realistic chemical models, rather than simplified systems. The use of realistic models, such as Rh-doped TiO₂, is essential for ensuring that the results are relevant to real-world applications.

Despite concerns about maintaining these computational advantages with increasingly complex systems, this work offers a vital benchmark for future quantum algorithm development. The challenging model of catalytic activity, nitrogen monoxide on rhodium-doped titanium dioxide, provides a key test case for assessing performance. The choice of NO adsorption on Rh-doped TiO₂ is particularly relevant because it exhibits strong correlation effects, multiple state crossings, and localized electronic states, all of which pose significant challenges for traditional computational methods. This new technique sharply reduces computational demands compared to existing methods, paving the way for more detailed simulations of chemical reactions and potentially accelerating materials discovery in the coming decade. The ability to accurately simulate catalytic processes could lead to the design of new catalysts with improved activity, selectivity, and stability. By demonstrating near-complete active space configuration interaction (CASCI) accuracy with fewer computational parameters, the adaptive, problem-tailored ansatz offers a strong advantage over traditional methods like state-averaged factorized unitary coupled cluster (SA-fUCCSD). CASCI is considered the ‘gold standard’ in quantum chemistry, providing highly accurate results but at a prohibitive computational cost. The research utilised an embedded cluster model of nitrogen monoxide adsorption on rhodium-doped titanium dioxide, a complex system exhibiting multiple electronic states, allowing direct comparison against established wavefunction techniques. The multiple electronic states arise from the different ways in which the electrons can be distributed within the molecule and the surface, leading to different potential energy surfaces and reaction pathways.

The researchers demonstrated that an adaptive, problem-tailored ansatz, SA-ADAPT, accurately models complex chemical processes with fewer computational parameters than existing methods. This is important because accurately simulating catalytic reactions, such as nitrogen monoxide adsorption on rhodium-doped titanium dioxide, is computationally demanding. SA-ADAPT achieved accuracy comparable to the ‘gold standard’ CASCI method while requiring significantly fewer operators, offering a more efficient approach to modelling these systems. The authors established a benchmark for evaluating quantum algorithms in chemistry using this challenging test case.