Silicon heterojunction solar cells have record-high open-circuit voltages but suffer from reduced short-circuit currents due in large part to parasitic absorption in the amorphous silicon, transparent conductive oxide (TCO), and metal layers. We previously identified and quantified visible and ultraviolet parasitic absorption in heterojunctions; here, we extend the analysis to infrared light in heterojunction solar cells with efficiencies exceeding 20%. An extensive experimental investigation of the TCO layers indicates that the rear layer serves as a crucial electrical contact between amorphous silicon and the metal reflector. If very transparent and at least 150 nm thick, the rear TCO layer also maximizes infrared response. An optical model that combines a ray-tracing algorithm and a thin-film simulator reveals why: parallel-polarized light arriving at the rear surface at oblique incidence excites surface plasmons in the metal reflector, and this parasitic absorption in the metal can exceed the absorption in the TCO layer itself. Thick TCO layers-or dielectric layers, in rear-passivated diffused-junction silicon solar cells-reduce the penetration of the evanescent waves to the metal, thereby increasing internal reflectance at the rear surface. With an optimized rear TCO layer, the front TCO dominates the infrared losses in heterojunction solar cells. As its thickness and carrier density are constrained by anti-reflection and lateral conduction requirements, the front TCO can be improved only by increasing its electron mobility. Cell results attest to the power of TCO optimization: With a high-mobility front TCO and a 150-nm-thick, highly transparent rear ITO layer, we recently reported a 4-cm 2 silicon heterojunction solar cell with an active-area short-circuit current density of nearly 39 mA/cm 2 and a certified efficiency of over 22%. V C 2013 American Institute of Physics. [http://dx