Author links open overlay panelPierre Verlinden 1, David L. Young 2, Gang Xiong 3, Matthew O. Reese 2, Lorelle M. Mansfield 2, Michael Powalla 4, Stefan Paetel 4, Ryan M. France 2, Philip T. Chiu 5,https://doi.org/10.1016/j.device.2023.100013Get rights and contentPhotovoltaics (PV) now produces the lowest-cost electricity in many parts of the world. Over three-quarters of the world's population lives in the sunbelt, and PV electricity is predicted, in many renewable energy scenarios, to become the majority global energy source by 2050. Device innovation and high-volume manufacturing have been central to the PV revolution. Continued research is important to increase efficiency, improve reliability, and decrease costs, all of which combine to make the increasingly compelling value proposition for PV power generation.PV device performance depends on the coupling of optical absorption, carrier transport, and interface control, fundamentals shared with a wide range of semiconductor devices and detectors. This perspective reviews recent advances and future directions in the three commercial-scale (yearly production greater than 1 GW) PV material technologies, as well as the III–V multijunction technology used for many satellite applications.Photovoltaics (PV), also known as solar cells, are now found everywhere—in utility plants; on roofs of homes and commercial buildings; on platforms at sea; in agricultural fields; on vehicles, buildings, drones, and backpacks; and, in their longest running application, providing power in space. Continuous device innovation has led to increased efficiency and improved reliability for multiple PV technologies. Confronted with an urgent need to deploy PV at multiterawatt (TW) scale over the next two decades to mitigate greenhouse gas emissions, PV device innovation takes on new urgency and impact. This perspective reviews recent progress in device design and performance for PV technologies that are currently in commercial production at greater than 1 GW/year or enabling significant space-based power generation—Si, CdTe, CIGS, and multijunction III–V—and looks ahead to the next 5 years. We also identify device-related topics requiring cross-cutting research and innovation.••140 years ago, inventor Charles Fritts made solar cells from selenium, hoping to offer an alternative to the coal-fired power plant that Thomas Edison built in New York City the year before.1 The 1%–2% efficient devices, Au on Se, were installed on a roof top in 1884 but obviously gained limited traction. The first practical Si solar cell was introduced in 1954 with an efficiency of ∼6%. Since then, photovoltaic devices based on several materials systems have moved to gigawatt (GW)-scale global annual production, and total installed global photovoltaics (PV) capacity is in excess of 1 terawatt (TW).PV is now the lowest-cost form of electricity in many parts of the world and is predicted, in many renewable energy scenarios,2 to become the majority energy source for the world by 2050. Although the 1 TW of installed generating capacity was a major global milestone, it is important to note that PV's contribution to worldwide electricity generation remains small: roughly 5%–6% for 2022. At the same time, PV has reached record levels of grid penetration in regions around the world, exceeding or approaching 20% of electricity for California, South Australia, and Hawaii.3Rapidly advancing device innovation and sophistication have been central to the growth of the PV industry in the past two decades. PV is a technology tha. Crystalline silicon devices represented ∼95% of the PV market in 2022. Their deployment requires inexpensive, large-area production methods of >1 TW/year projected by 2028 (or >5 billion m2 per year) and operation in a harsh outdoor environment for more than 25 years with <0.5% annual degradation. The device structure evolved rapidly from a planar n+/p diffused phosphorus junction on a p-type wafer to the incorporation of a back surface field (BSF) to form an n+/p/p+ structure with diffused aluminum forming high-low junctions and the back contact (Figure 1A). Front surface texturization and an anti-reflection coating increased the current density.This basic structure survived until about 2015 with a multitude of incremental improvements (screen printed contacts, SiNx passivation and anti-reflection layers, selective emitters, and high-lifetime multicrystalline wafers) to raise the efficiency from 6% to ∼22% and to reduce the cost of manufacturing from about US$1,000/W in 1960 to less than US$1/W in 2015. Over the last five decades, the learning curve model has shown that the cost of manufacturing silicon PV modules has decreased with an average learning rate of 24% for every doubling of the cumulative production volume, with some obvious “bumps” when the global production is limited by supply constraints followed by a rapid decrease when the supply chain recovers. Efficiency and economies of scale are import.