Numerical Aspects of the Three-Dimensional Feature-Scale Simulation of Silicon-Nanowire Field-Effect Sensors for DNA Detection Clemens Heitzinger and Gerhard Klimeck Network for Computational Nanotechnology, School of Electrical and Computer Engineering, Purdue University, West Lafayette, IN 47907, USA E-mail: ClemensH@Purdue.edu I NTRODUCTION AND MOTIVATION In recent months experimental silicon-nanowire field-effect biosensors were built and their function- ing was verified [1], [2]. These devices consist of a silicon-nanowire core, an enveloping silicon-oxide, and a surface receptor molecule (cf. Fig. 1). When a biomolecule attaches to the surface receptor and this attachment results in a sufficiently different charge distribution, the change in current flow through the nanowire enables detection. These sensors provide perfect selectivity since the possibility of a binding being established between two biomolecules (a protein and an antibody, for ex- ample) is equivalent to having a biological function. Furthermore extremely high detection sensitivity in the pg/ml regime has been reported [1]. It seems feasible that these devices can sense a huge array of biomolecules, and notable application areas are the detection of cancer markers and DNA fragments. SIMULATION METHOD In this work we investigate the numerical as- pects and challenges of three-dimensional feature- scale simulations at the example of three states of a nanoscale DNA sensor in aqueous solution. In the first state nothing is attached, in the sec- ond state one chain of the DNA fragment (5’- D(CGTGAATTCACG)-3’) is attached, and in the third state the whole dodecamer is attached (see Fig. 1, Fig. 2, Fig. 3). After determining the partial charges on the DNA fragment, their distribution was used to obtain the electrostatic potential by solving the 3D Poisson equation. Charge transport was simulated using a 3D self-consistent NEGF simulator [3]. RESULTS AND CONCLUSIONS Simulator timings are shown in Fig. 4, the current-voltage characteristics in Fig. 5, and the potential in Fig. 6. The characteristics imply that the differences in current between the three states allow to discern if a functional device was produced (i.e., a single-stranded fragment is attached) and if a second strand is attached to the first. The simulations show that the length of the linker is a critical device parameter. The detection of larger and only moderately charged molecules will be correspondingly more difficult. The calculation of the electrostatic potential around the molecules and in the nanowire neces- sitates the use of sparse-matrix representations and algorithms to achieve good resolution within modest memory requirements. Transport simulations using the NEGF formalism benefit from parallelization. (Readers will be able to run simulations on vari- ous structures online at http://www.nanohub.org.) ACKNOWLEDGMENT This material is based upon work supported by the National Science Foundation under Grant No. EEC-0228390, the Indiana 21st Century fund, and the Semiconductor Research Corporation. REFERENCES [1] G. Zheng, F. Patolsky, Y. Cui, W.U. Wang, and C.M. Lieber. Nature Biotechnology, 23:1294–1301, October 2005. [2] C. Yang, Z. Zhong, and C.M. Lieber. Science, 310:1304– 1307, November 2005. [3] J. Wang, E. Polizzi, and M. Lundstrom. In IEDM 2003, pages 695–698, Washington DC, USA, December 2003. 11th International Workshop on Computational Electronics TU Wien, 25-27 May 2006 ISBN 3-901578-16-1 169