Hamiltonian simulation is a fundamental application of quantum computing, particularly in quantum chemistry, where solving the electronic structure problem requires efficiently representing electron repulsion integrals (ERI). This thesis analyzes the computational complexity of Hamiltonian simulation using the Linear Combination ofUnitaries (LCU) algorithm and Tensor Hypercontraction (THC), a factorization method that reduces the cost of ERI representation. By decomposing the Coulomb tensor via Interpolative Separable Density Fitting (ISDF) and mapping the THC representation to an LCU form using the Jordan-Wigner transformation, we achieve a more efficient block encoding of the Hamiltonian in a basis of N molecular orbitals. This enables quantum simulation with an asymptotic Toffoli complexity of O(Nl/e) and a space complexity of O(N), where e = 0.0016 Hartree is the target accuracy and l denotes the 1-norm of the coefficients of the electronic Hamiltonian in the THC representation. Beyond Hamiltonian simulation, we explore how Quantum Signal Processing (QSP) and Quantum Singular Value Transformation (QSVT) leverage block encoding for eigenvalue transformations and time evolution, extending its utility to a broader class of quantum algorithms. Despite these advancements, large-scale quantum chemistry simulations are far from experimental feasibility, for instance, simulating the FeMoCo molecule in an active space model under surface code constraints would require billions of Toffoli gates, millions of physical qubits and several days of runtime using error-corrected qubits; and that is assuming it ever becomes feasible. This work contributes to the effort of making quantum chemistry simulations more computationally viable by investigating how LCU and THC can be integrated to reduce costs and maintain chemical accuracy, while exploring different methods and their combinations to identify key challenges in quantum chemistry simulations.
Quantum chemistry models atoms and molecules using quantum mechanics, with the electronic structure problem at its core. This problem, which determines the energy and properties of electrons in a molecule, is exponentially hard to solve exactly on classical computers. To approximate solutions, effective models introduce layers of simplifications. However, these approximations can fail for strongly correlated systems, necessitating alternative approaches. Quantum computers, capable of processing quantum information natively, offer a promising path to simulate molecular systems beyond classical reach, with potential applications in catalysis, materials science, and photochemistry. Despite rapid advancements in quantum hardware, practical quantum computations remain challenging due to limited resources. This thesis optimizes quantum resource usage in computational chemistry, introducing novel embedding techniques and exploring applications beyond ground-state energies. Chapter 1 establishes the foundation by reviewing quantum chemistry methods, from Hartree-Fock and Density Functional Theory (DFT) to active space approaches. It also provides an introduction to quantum computing for chemistry, focusing on near-term algorithms. The following chapters present novel methods aimed at reducing quantum resource requirements and tackling problems classical methods struggle with. Chapter 2 introduces a technique to reduce the
λ-norm of the electronic structure Hamiltonian, a key measure of quantum resource demand. By transforming molecular orbitals, this method enhances sampling efficiency in Noisy Intermediate-Scale Quantum (NISQ) algorithms and reduces runtime in Fault-Tolerant Quantum Computing (FTQC). Localized orbitals outperform canonical ones, and an orbital-optimization scheme is proposed to minimize the
λ-norm. Benchmarks on hydrogen chains, FeMoco, and Ruthenium complexes demonstrate its effectiveness. Chapter 3 presents the FragPT2 framework for simulating large systems with multiple interacting fragments. This method embeds each fragment in the mean field of others and applies multi-reference perturbation theory to capture inter-fragment correlations such as dispersion and charge transfer efficiently. The approach reduces computational overhead and enables scalable simulations, as shown in applications to \ce{N2} dimers, aromatic dimers, and butadiene. Chapter 4 focuses on photochemical systems, where multiple electronic states must be treated equally. The state-averaged orbital-optimized variational quantum eigensolver (SAOOVQE) is extended with analytical energy gradients and non-adiabatic couplings, enabling geometry optimizations and simulations of non-adiabatic dynamics on quantum hardware. This is demonstrated by locating the conical intersection in formaldimine, highlighting SAOOVQE’s potential for studying photoisomerization in biomolecules. Building on this, Chapter 5 develops a quantum algorithm to detect conical intersections, where the Born-Oppenheimer approximation breaks down. These intersections are characterized by a discrete Berry phase (0 or
π), representing their topological nature. The algorithm extracts this phase as a single bit of information, ensuring resilience to noise. Analytical proof and numerical validation using a toy molecular model illustrate its effectiveness for studying light-perception biochemistry and related photochemical processes. Finally, Chapter 6 explores machine learning density functionals trained on noisy quantum data. While DFT is computationally efficient, its accuracy depends on approximated exchange-correlation functionals. This work proposes a neural network-based approach to learn the exact functional from quantum data, showing that training on noisy outputs from NISQ devices can mitigate sampling errors. Applications to the Fermi-Hubbard model and other systems demonstrate its potential to generalize and provide accurate results under noise. This thesis advances quantum computational chemistry by optimizing quantum resource usage, developing new embedding strategies, and expanding the scope of quantum simulations beyond ground-state properties. These contributions pave the way for practical quantum algorithms that could revolutionize molecular simulations in the near future.
We develop a workflow decomposing a molecule into fragments via quantum embedding and simulating them with a heterogeneous quantum-classical (HQC) method. We sample fragment electronic configurations on two 156-qubit quantum processors (ibm
_cleveland, ibm
_kobe), using up to 94 qubits, running 21,006 circuits for over 239 hours, collecting
3.0⋅109 measurement outcomes -- the most resource-intensive HQC computation for quantum chemistry to date. We compute fragment wavefunctions via optimized subspace diagonalization on supercomputers Fugaku, Miyabi-G, and ROQUO, achieving 72.5
% parallel efficiency with scalable distributed linear algebra kernels. We simulate two protein-ligand complexes spanning dispersion- and electrostatics-dominated regimes (11,608, 12,635 atoms), demonstrate
>40× increase in system size and up to
210× improvement in accuracy over previous state-of-the-art, with HQC matching coupled-cluster (CCSD) accuracy in fragment energies. We present the first HQC protein-ligand binding prediction using a mixed-basis set and an automated end-to-end workflow enabling practical HQC calculations of large protein systems.
Quantum computers must operate in concert with classical computers to deliver on the promise of quantum advantage for practical problems. To achieve that, it is important to understand how quantum and classical computing can interact together, and how one can characterize the scalability and efficiency of hybrid quantum–classical workflows. So far, early experiments with quantum-centric supercomputing workflows have been limited in scale and complexity. Here, we use a Heron quantum processor deployed on premises with the entire supercomputer Fugaku to perform the largest computation of electronic structure involving quantum and classical high-performance computing. We design a closed-loop workflow between the quantum processors and 152,064 classical nodes of Fugaku, to approximate the electronic structure of chemistry models beyond the reach of exact diagonalization, with accuracy comparable to some all-classical approximation methods. Our work pushes the limits of the integration of quantum and classical high-performance computing, showcasing computational resource orchestration at the largest scale possible for current classical supercomputers.
Reducing the complexity of quantum algorithms to treat quantum chemistry problems is essential to demonstrate an eventual quantum advantage of noisy-intermediate scale quantum devices over their classical counterpart. Significant improvements have been made recently to simulate the time-evolution operator $U(t)={e}^{i\stackrel{\ifmmode \hat{}\else \^{}\fi{}}{\mathcal{H}}t}$, where $\stackrel{\ifmmode \hat{}\else \^{}\fi{}}{\mathcal{H}}$ is the electronic structure Hamiltonian, or to simulate $\stackrel{\ifmmode \hat{}\else \^{}\fi{}}{\mathcal{H}}$ directly (when written as a linear combination of unitaries) by using block encoding or qubitization techniques. A fundamental measure quantifying the practical implementation complexity of these quantum algorithms is the so-called 1-norm of the qubit representation of the Hamiltonian, which can be reduced by writing the Hamiltonian in factorized or tensor-hypercontracted forms, for instance. In this paper, we investigate the effect of classical preoptimization of the electronic structure Hamiltonian representation, via single-particle basis transformation, on the 1-norm. Specifically, we employ several localization schemes and benchmark the 1-norm of several systems of different sizes (number of atoms and active space sizes). We also derive a formula for the 1-norm as a function of the electronic integrals and use this quantity as a cost function for an orbital-optimization scheme that improves over localization schemes. This paper gives more insights about the importance of the 1-norm in quantum computing for quantum chemistry and provides simple ways of decreasing its value to reduce the complexity of quantum algorithms.
The most advanced techniques using fault-tolerant quantum computers to estimate the ground-state energy of a chemical Hamiltonian involve compression of the Coulomb operator through tensor factorizations, enabling efficient block encodings of the Hamiltonian. A natural challenge of these methods is the degree to which block-encoding costs can be reduced. We address this challenge through the technique of spectral amplification, which magnifies the spectrum of the low-energy states of Hamiltonians that can be expressed as sums of squares. Spectral amplification enables estimating ground-state energies with significantly improved cost scaling in the block encoding normalization factor Λ to just 2 Λ E gap , where E gap ≪ Λ is the lowest energy of the sum-of-squares Hamiltonian. To achieve this, we show that sum-of-squares representations of the electronic structure Hamiltonian are efficiently computable by a family of classical simulation techniques that approximate the ground-state energy from below. In order to further optimize, we also develop a novel factorization that provides a trade-off between the two leading Coulomb integral factorization schemes—namely, double factorization and tensor hypercontraction—that when combined with spectral amplification yields a factor of 4 to 195 speedup over the state of the art in ground-state energy estimation for models of iron-sulfur complexes and a CO 2 -fixation catalyst.