IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 55, NO. 9, SEPTEMBER 2008 2287 Impact of Donor Concentration, Electric Field, and Temperature Effects on the Leakage Current in Germanium p+/n Junctions Geert Eneman, Maxime Wiot, Antoine Brugère, Oriol Sicart I Casain, Sushant Sonde, David P. Brunco, Brice De Jaeger, Alessandra Satta, Geert Hellings, Kristin De Meyer, Senior Member, IEEE, Cor Claeys, Senior Member, IEEE, Marc Meuris, Marc M. Heyns, and Eddy Simoen Abstract—This paper presents an analysis of junction leak- age in heavily doped p+/n germanium junctions, targeted for short-channel transistor fabrication. There exists an opti- mal p+/n junction condition, with a doping concentration of 1 × 10 17 –5 × 10 17 cm −3 , where the area-leakage-current density is minimal. Use of a halo-implant condition optimized for our 125-nm gate-length pMOS devices shows less than one decade higher area leakage than the optimal p+/n junction. For even higher doping levels, the leakage density increases strongly. There- fore, careful optimization of p+/n junctions is needed for deca- nanometer germanium transistors. The junction leakage shows good agreement with electrical simulations, although for some implant conditions, more adequate implant models are required. Finally, it is shown that the area-junction static-power consump- tion for the best junctions remains below the power-density speci- fications for high-performance applications. Index Terms—Germanium, halo implant, leakage current, MOSFETs, p+/n junction, trap-assisted tunneling (TAT). I. INTRODUCTION O NE OF THE concerns in the development of Ge-channel MOS devices is the small band gap (E g = 0.67 eV), which can give rise to Band-to-Band Tunneling (BTBT) when sufficiently high electric fields exist at the drain junction of the transistor [1], [2]. High electric fields can be expected in deep- submicrometer devices, owing to the high substrate-doping levels typically used to control the short-channel effects. It is common practice to employ a so-called halo or pocket implan- tation, giving rise to a high doping concentration at the junction Manuscript received December 26, 2007; revised April 9, 2008. The review of this paper was arranged by Editor J. Cressler. G. Eneman is with the Interuniversity MicroElectronics Center (IMEC), 3001 Leuven, Belgium, with the Fund for Scientific Research-Flanders, 1000 Brussels, Belgium, and also with the Electrical Engineering Department, INSYS Division, Katholieke Universiteit Leuven, 3001 Leuven, Belgium (e-mail: eneman@imec.be). M. Wiot, A. Brugère, O. Sicart I Casain, S. Sonde, B. De Jaeger, A. Satta, M. Meuris, M. M. Heyns, and E. Simoen are with the Interuni- versity MicroElectronics Center (IMEC), 3001 Leuven, Belgium (e-mail: maxime.wiot@gmail.com). D. P. Brunco is an Intel Assignee at the Interuniversity MicroElectronics Center (IMEC), 3001 Leuven, Belgium. G. Hellings, K. De Meyer, and C. Claeys are with the Interuniversity MicroElectronics Center (IMEC), 3001 Leuven, Belgium, and also with the Electrical Engineering Department, INSYS Division, Katholieke Universiteit Leuven, 3001 Leuven, Belgium. Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TED.2008.927660 [3]. In order to suppress the phenomenon, dedicated architec- tures using thin Ge epitaxial layers have been proposed [4], [5]. BTBT or Zener tunneling is in fact an “old problem” in germanium p-n junctions—one of the first reports was al- ready published in 1951 [6]. Besides direct tunneling across the germanium band gap, deep-level generation–recombination centers may also assist in the transition of electrons from the valence to the conduction band [7], so-called Trap-Assisted Tunneling (TAT). For the case of silicon p-n junctions, the electric-field (F ) dependence can be described following the approach of Hurkx et al. [8], [9], resulting in an expression often used in device simulations R trap (x)=[1 + Γ(x)] R SHR (x) (1) where R trap is the field-assisted recombination rate (in per cubic centimeter per second) at a distance x from a one- sided junction and R SRH is the Shockley–Read–Hall (SRH) recombination rate through phonon-assisted carrier processes via a deep trap. The electric-field-enhancement factor Γ(x) corresponds to [8], [9] Γ= ΔE kT 1  o exp  ΔE kT u − 4 3 √ 2m ∗ (ΔE) 3 q|F | u 3/2  du (2) where F is the magnitude of the electric field, kT is the thermal energy, q is the absolute value of the electron charge,  is the reduced Planck constant, and m ∗ is the effective mass of the carriers. ΔE is related to the trap level and equals E g /2 for midgap states. The integration variable u is related to the energy of the carriers after tunneling, as is detailed in [8]. Having this in mind, the goal of the work reported in this paper is to evaluate the leakage in p+/n junctions fabricated with different halo doping densities, i.e., with different maxi- mum electric field F max at the junction. Ideally, one expects a dominance of standard SRH generation at low donor concen- trations N D , going over to TAT at intermediate concentrations and to BTBT at still higher concentrations. This transition point will also depend on the bias condition and temperature. As will be shown, for the Ge p+/n junctions under study, an optimum peak N D exists in the range of 1–5 × 10 17 cm −3 , where the leakage-current density at a fixed V R = 0.5 V is minimum. Be- yond that point, TAT governs both the area (J A ) and perimeter 0018-9383/$25.00 © 2008 IEEE