Silicon LEDs in FinFET technology G. Piccolo 1 , P.I. Kuindersma 3 , L-A. Ragnarsson 2 , R.J.E. Hueting 1 , N. Collaert 2 , and J. Schmitz 1,* 1 MESA+ Institute for Nanotechnology, University of Twente, Enschede, The Netherlands 2 imec, Leuven, Belgium. 3 NXP Semiconductors (now at Eindhoven University of Technology), Eindhoven, Netherlands * j.schmitz@utwente.nl Abstract—We present what to our best knowledge is the first forward operating silicon light-emitting diode (LED) in fin- FET technology. The results show near-infrared (NIR) emission around 1100 nm caused by band-to-band light emission in the silicon which is uniformly distributed across the lowly doped active light-emitting area. We also propose further improvements to exploit the full potential of this structure. Index Terms—Silicon, Silicon on insulator technology, Electro- luminescence, Elemental semiconductors, Light emitting diodes, p-i-n diodes, FinFETs, LED, Carrier injectors, Integrated op- tics, Near-infrared light emission, Infrared light sources, Silicon Photonics, Nanometric devices. I. I NTRODUCTION Efficient light emission from silicon-based devices has been presented as one of the holy grails in electrical engineering [1]. The literature reports on a wide range of potential candidate light emitters that may be integrated with microelectronics. Most effort is spent on integrating optical active material on silicon, that is mainly used as passive carrier. For a review of these, see e.g. [2], [3]. If we consider the criteria to target a successful integration in state-of-the-art VLSI platforms, i.e. a full compliance to CMOS fabrication standards and an operative voltage conform to the supply range of modern chips (or even better suitable for low power applications), many of the approaches reported in [2], [3] are not optimal. Band- edge emission from silicon, instead, can be stimulated at low power with relatively simple device architectures: therefore we present a silicon LED based on a forward-biased p-i-n diode structure. The choice of silicon as an active (light-emitting) material is less intuitive. The radiative lifetime for electron-hole recom- bination in silicon is relatively long compared to non-radiative lifetimes, and therefore the probability of non-radiative re- combination is higher. Shockley-Read-Hall (SRH) recombi- nation will dominate at low carrier injection levels, while Auger recombination at high injection (above ˜10 18 carriers per cm -3 ). Proper tuning of the injection level, combined with high quality material to limit SRH, allows to benefit from the relatively low surface recombination of the interface silicon- thermal oxide, resulting in a potentially sufficient efficiency. Several groups [4], [5] have been predicting a theoretical maximum of 20 % internal efficicency for silicon LEDs. In practice, the efficiency record, to the best of our knowledge, was achieved by Green et al. [6] with a reported 0.92% external quantum efficiency on forward-biased silicon LEDs. ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ + n + p i L W Anode Cathode n-gate p-gate fins Fig. 1. Schematic topview of the LED. The device is a p-i-n diode with an intrinsic (active) region of L × W . The intrinsic region is connected to the extensions via a series of gated fins, whose number depends on the actual dimensions of the device (dimensions and pitch of the fins are fixed). The actual junction on each side falls beneath the gate. The figure is not to scale. Fig. 2. Schematic cross-section of the LED in the sense of the current, along one fin. The device is a p-i-n diode gates at the two junction sides. The dashed lines delimit the fin region. The figure is not to scale. Yet, the LED presented by Green et al. has an active (light emitting) volume in the order of mm 3 , whereas much higher granularity is necessary both to reduce costs and maximize the out-coupling of the light in integrated waveguides. Compact, efficient LEDs has been previously fabricated on silicon-on- insulator (SOI) with an active volume of 24 μm 3 [7]. In such miniaturized configurations, surface recombination and carrier injector design come into play, as demonstrated in recent work from our group, where studies of injector scaling [8], [9] indicate that a smaller injector may lead to better quantum efficiency. When scaling down to nano-injectors, it is imperative to ensure a smooth conduction (i.e. no interfaces in the current path) as well as to control properly the quality of the surfaces of the injectors, to limit surface recombination. To 978-1-4799-4376-0/14/$31.00 ©2014 IEEE 274