Abstract
Octree meshes provide an efficient framework for hierarchical mesh refinement. However, transitions between refinement levels introduce hanging nodes and generate polyhedral cells with varying boundary configurations. Exploiting rotational and reflectional symmetries reduces the admissible balanced-octree patterns to a limited set of unique configurations. For each configuration, a parametric parent element is constructed, and scaled boundary shape functions are derived using a dual scaled boundary approach that combines polygonal face interpolation with a second scaling into the polyhedral volume. The shape functions and their derivatives are precomputed at the integration points and reused for all corresponding physical cells, such that the underlying eigenvalue problem is solved only once per parent element. The interpolation functions are further analyzed in the parent-element space using Duffy-type transformations. Their monomial structure provides the basis for modifying higher-order eigenvalues, reducing the required quadrature order while preserving linear completeness of the shape functions. The resulting elements are incorporated into a total Lagrangian finite-strain formulation. Numerical examples, including a nonlinear patch test, Cook's membrane, a heterogeneous cube with a spherical inclusion and a locally refined coin-like geometry, demonstrate consistency and convergence of the formulation and its applicability to uniform and locally refined octree discretizations, and heterogeneous mesh domains.