This paper investigates energy filtering in silicon nanowires and nanosheets by resonant electron tunneling through a geometric superlattice. A geometric superlattice is any kind of periodic geometric feature along the transport direction of the nanowire or nanosheet. Multivalley quantum-transport simulations are used to demonstrate the manifestation of minibands and minibandgaps in the transmission spectra of such a superlattice. We find that the presence of different valleys in the conduction band of silicon favors a nanowire with a rectangular cross section for effective energy filtering. The obtained energy filter can consequently be used in the source extension of a field-effect transistor to prevent high-energy electrons from contributing to the leakage current. Self-consistent Schrodinger-Poisson simulations in the ballistic limit show minimum subthreshold swings of 6 mV/decade for geometric superlattices with indentations. The obtained theoretical performance metrics for the simulated devices are compared with conventional III-V superlatticeFETs and TunnelFETs. The adaptation of the quantum transmitting boundary method to the finite-element simulation of 3-D structures with anisotropic effective mass is presented in Appendixes A and B. Our results bare relevance in the search for steep-slope transistor alternatives which are compatible with the silicon industry and can overcome the power-consumption bottleneck inherent to standard CMOS technologies. Published by AIP Publishing.