In order to study a possibility of superfluidity in trapped atomic Fermi gases loaded on optical lattices, we implement an exact diagonalization code for the trapped Hubbard model on the Earth Simulator. Comparing two diagonalization algorithms, we find that the performance of the preconditioned conjugate gradient (PCG) method is 1.5 times superior to the conventional Lanczos one since the PCG method can conceal the communication overhead much more efficiently. Consequently, the PCG method shows 16.447 TFlops (50.2% of the peak) on 512 nodes. On the other hand, we succeed in solving a 159-billion-dimensional matrix by using the conventional Lanczos method. To our knowledge, this dimension is a world-record. Numerical results reveal that an unconventional type of superfluidity specific to the confined system develops under repulsive interaction.
In order to study a possibility of superfluidity in trapped atomic Fermi gases loaded on optical lattices, we implement an exact diagonalization code for the trapped Hubbard model on the Earth Simulator. Comparing two diagonalization algorithms, we find that the performance of the preconditioned conjugate gradient (PCG) method is 1.5 times superior to the conventional Lanczos one since the PCG method can conceal the communication overhead much more efficiently. Consequently, the PCG method shows 16.447 TFlops (50.2% of the peak) on 512 nodes. On the other hand, we succeed in solving a 159-billion-dimensional matrix by using the conventional Lanczos method. To our knowledge, this dimension is a world-record. Numerical results reveal that an unconventional type of superfluidity specific to the confined system develops under repulsive interaction.
Abstract In order to explore a possibility of superfluidity in confined strongly-correlated fermion systems, e.g., nano-scale cuprate High-Tc superconductors and atomic Fermi gases loaded on optical lattice, we imple-ment an exact diagonalization code for their mathematical model, i.e., a trapped Hubbard model on the Earth Simulator. We compare two diagonalization algorithms, the traditional Lanczos method and a new algorithm, the preconditioned conjugate gradient (PCG) method, and find that when using the PCG the total CPU time can be reduced to 1/3 ~ 1/5 compared to the former one since the convergence can be dramatically improved by choosing a good preconditioner and the communication overhead is much more efficiently concealed in the PCG method. Consequently, such a performance improvement enables us to do systematic studies for sev-eral parameters. Numerical simulation results reveal that an unconventional type of pairing specific to the confined system, which may cause superfluidity, develops under a strong repulsive interaction.
Over the past few decades, quantum simulation with ultracold atoms in optical lattices has proven to be a useful tool for understanding many-body Hamiltonians.One of its primary directions is the experimental study of the Fermi Hubbard Model (FHM), which aims to capture the fundamental properties of cuprate superconductors and has produced a wide range of intriguing phenomena, including the metal-to-Mott insulator transition, antiferromagnetism, strange metallicity, pseudogap physics, and is frequently used in the study of superconductivity. In recent years, new kinds of quantum simulators have emerged, and in particular we focus on two of them: 1) quantum simulators that study high symmetry fermions in optical lattices, and 2) optical tweezer arrays which provide a bottom-up approach to engineering the FHM. The first class of emerging quantum simulators utilizes alkaline-earth-like atoms (AEA’s) – which have large nuclear spin I (such as 173Yb and 87Sr) – to engineer the SU(N = 2I +1) symmetric FHM, where N = 2,..., 10. The second class corresponds to quantum simulators based on optical tweezer arrays which combine deterministic single tweezer ground state loading and tunneling coupled sites. Both classes of quantum simulators are interesting on their own. For example, the SU(N) FHM is generating a lot of interest due to its connections to multiorbital solid state systems and theoretical predictions of rich phase diagrams which exhibit exotic phases of matter, including unusual ones like spin liquids. On the other hand, optical tweezer arrays have gained a lot of attention since they can be used to study arbitrary geometries and access very low-entropy samples. This thesis reports on the numerical study of both quantum simulation architectures. Results of the SU(N) FHM at finite temperature, for different fillings and interaction strengths that span the non-interacting to the strongly interacting limit are obtained using Determinant Quantum Monte Carlo, Exact Diagonalization, and Numerical Linked Cluster Expansion. In particular, we focus on how thermodynamic and magnetic observables, such as the number of on-site pairs, energy, entropy, spin correlations, and structure factors depend on N, the interaction strength U/t, temperature T/t, and the anisotropy of tunneling rates. Our results demonstrate that in a homogeneous square lattice with one particle per site on average, thermodynamic observables as a function of temperature obey a universal scaling with N, and that short-range antiferromagnetic correlations are stronger for larger N and in lower dimensions. In addition, we present theory-experiment comparisons where possible, where we perform thermometry and provide a precise characterization of the equation of state of the SU(N) FHM. Our results emerge as a tool to perform thermometry in experiments from theory-experiment comparison and also provide guidance for future experiments with AEAs in optical lattices. Additionally, we present calculations using numerically precise discrete variable representation methods for two-dimensional stroboscopic tweezer arrays and compare the outcomes with experimental data. We quantify the effects of stroboscopic potentials on Hubbard parameters like the interaction strength U and the tunneling t in optical tweezer arrays and illustrate how heating from the stroboscopic potential relies on strobe frequency. Our calculations enable evaluation and optimization of two-dimensional tweezer array experiments. An important milestone of the results of this thesis corresponds to the record temperatures and entropies achieved with these quantum simulation architectures: 1) the coldest fermions ever created in nature in absolute temperature and in cold atoms, and 2) the lowest entropy per particle fermions ever created, albeit in a so-far small system.
We investigate a possibility of superfluidity in a trapped gas of Fermi atoms with a repulsive interaction in the presence of an optical lattice. Applying the exact diagonalization method to a one-dimensional Hubbard model including the trap potential, we find that, when the strength of the repulsive interaction exceeds a critical value, the binding energy of two Fermi atoms becomes negative below the half-filling case, indicating that an attractive interaction effectively works between Fermi atoms. In this case, a "Mott insulating core" appears in the center of the trap, where each site is occupied by one atom. The Cooper-pair correlation strongly develops between atoms in the left- and right-hand sides of this core.
One of the challenging features of studying model Hamiltonians with cold atoms in optical lattices is the presence of spatial inhomogeneities induced by the confining potential, which results in the coexistence of different phases.This paper presents quantum Monte Carlo results comparing methods for confining fermions in two dimensions, including conventional diagonal confinement, a recently proposed ``off-diagonal confinement'', as well as a trap which produces uniform density in the lattice. At constant entropy and for currently accessible temperatures, we show that (1) diagonal confinement results in the strongest signature of magnetic order, primarily because of its judicious use of entropy sinks at the trap edge and that (2) for d-wave pairing, a trap with uniform density is optimal and can be effectively implemented via off-diagonal confinement. This feature is important to any prospective search for superconductivity in optical lattices.
Fermionic atoms in optical lattices provide a native implementation of Fermi-Hubbard (FH) models that can be used as analog quantum simulators of many-body fermionic systems. Recent experimental advances include the time-dependent local control of chemical potentials and tunnelings, and thus enable us to operate this platform digitally as a programmable quantum simulator. Here, we explore these opportunities and develop ground-state preparation algorithms for different fermionic models, based on the ability to implement both single-particle and many-body high-fidelity fermionic gates, as provided by the native FH Hamiltonian. In particular, we first design variational precompiled quantum circuits to prepare the ground state of the natively implemented FH model, with significant speedups relative to competing adiabatic protocols. Besides, the versatility of this variational approach enables us to target extended FH models, i.e., including terms that are not natively realized on the platform. As an illustration, we include next-nearest-neighbor tunnelings at finite dopings, relevant in the context of d -wave superconductivity. Furthermore, we discuss how to approximate the imaginary-time evolution using variational fermionic circuits, both as an alternative state-preparation strategy and as a subroutine for the quantum Lanczos algorithm to further improve the energy estimation. We benchmark our protocols for ladder geometries, though they can be readily applied to two-dimensional experimental setups to address regimes beyond the capabilities of current classical methods. These results pave the way for more efficient and comprehensive explorations of relevant many-body phases with existing programmable fermionic quantum simulators.
We investigate the one-dimensional Hubbard model with a confining potential, which may describe cold fermionic atoms trapped in an optical lattice. Combining the variational Monte Carlo simulations with the new stochastic reconfiguration scheme proposed by Sorella, we present an efficient method to systematically treat the ground state properties of the confined system with a site-dependent potential. By taking into account intersite correlations as well as site-dependent on-site correlations, we are able to describe the coexistence of the metallic and Mott insulating regions, which is consistent with other numerical results. Several possible improvements of the trial states are also addressed.