Numerical Block-Diagonalization and Linked Cluster Expansion for Deriving Effective Hamiltonians: Applications to Spin Excitations (2505.11167v1)

Abstract: We present a non-perturbative framework for deriving effective Hamiltonians that describe low-energy excitations in quantum many-body systems. The method combines block diagonalization based on the Cederbaum--Schirmer--Meyer transformation with the numerical linked-cluster (NLC) expansion. A key feature of the approach is a variational criterion that uniquely determines the unitary transformation by minimizing the transformation of the state basis within the low-energy subspace. This criterion also provides a robust guideline for selecting relevant eigenstates, even in the presence of avoided level crossing and mixing induced by particle-number-nonconserving interactions. We demonstrate the method in two quantum spin models: the one-dimensional transverse-field Ising model and the two-dimensional Shastry--Sutherland model, relevant to SrCu$_2$(BO$_3$)$_2$. In both cases, the derived effective Hamiltonians faithfully reproduce the structure and dynamics of magnon and triplon excitations, including the emergence of topological band structures. The block diagonalization enables quantum fluctuations to be incorporated non-perturbatively, while the NLC expansion systematically accounts for finite-size corrections from larger clusters. This approach naturally generates long-range effective interactions near criticality, even when the original Hamiltonian includes only short-range couplings. The proposed framework provides a general and computationally feasible tool for constructing physically meaningful effective models across a broad range of quantum many-body systems.