Solution growth of metal-organic complex CuTCNQ in small dimension interconnect structures A. Demolliens a , Ch. Muller b,n , R. M¨ uller c , Ch. Turquat a , L. Goux c , D. Deleruyelle b , D.J. Wouters c a im2np, Institut Mate´riaux, Microe´lectronique et Nanosciences de Provence, UMR CNRS 6242, Universite´ du Sud Toulon Var, BP 20132, F-83957 La Garde Cedex, France b im2np, Institut Mate´riaux Microe´lectronique et Nanosciences de Provence, UMR CNRS 6242, Aix-Marseille Universite´, IMT Technopˆ ole de Chˆ ateau Gombert, F-13451 Marseille Cedex 20, France c imec, Interuniversity MicroElectronics Center vzw., Kapeldreef 75, B-3001 Leuven, Belgium article info Article history: Received 27 April 2010 Received in revised form 12 July 2010 Accepted 5 August 2010 Communicated by M. Tischler Available online 14 August 2010 Keywords: A2. Growth for solutions B1. Charge-transfer complex B2. Semiconducting materials B3. Resistive switching memories abstract In this paper, we report two different elaboration routes to grow metal-organic complex CuTCNQ in liquid phase within small interconnect structures (i.e. via holes opened in SiO 2 /SiC stack). The basic common idea relies on the formation of CuTCNQ material from the partial corrosion of a Cu bottom electrode by a TCNQ-based solution. The two solution growth methods are compared in terms of (i) via holes filling; (ii) local microstructure of CuTCNQ complex and (iii) quality of interface between CuTCNQ and copper metallic electrode. In the first route, in the reaction of the substrate with a TCNQ/copper salt solution in acetonitrile/toluene, a rapid formation of porous CuTCNQ complex is observed with an over- growth outside interconnect structures and many voids within via holes and at the interface with Cu layer. In contrast to this ‘‘mushroom-like’’ growth, the reaction of the substrate with a TCNQ solution in acetonitrile/2-butanone results in a ‘‘crystal-like’’ dense CuTCNQ complex within via holes and a CuTCNQ/Cu interface free of voids. In the latter case, satisfactory electrical performances are expected for future resistive switching memory devices. & 2010 Elsevier B.V. All rights reserved. 1. Introduction The demand for non-volatile memories is rapidly growing due to increase of nomad applications. To overcome the limitations of Flash memories below the 22 nm technological node, new concepts no longer based on charge storage are widely investi- gated. Among emerging technologies, resistive switching mem- ories (so-called RRAM) integrating either binary oxides or metal-organic complexes appear as promising candidates [1–4]. Their basic metal/insulator/metal (MIM) structures undergo a reversible change in resistance by applying a voltage or current sweep. MIM structures are highly scalable and can be integrated into back-end of line (BEOL) in small dimension interconnect structures, allowing high density storage with possibly multi- levels three-dimensional architectures. The main criteria required for these devices are (i) good discrimination between high and low resistance states (i.e. large memory window); (ii) high endurance; (iii) long term retention and (iv) low cost integration. The metal-organic complex called CuTCNQ for copper-7,7 0 ,8,8 0 - tetracyanoquinodimethane fulfills the previous requirements. Its resistance change was first reported by Potember et al. in 1979 [5], who attributed the switching to a reversible charge transfer between CuTCNQ complex and neutral Cu and TCNQ 0 . In 1982, Kamitsos et al. [6] studied the mechanism of electrical switching in CuTCNQ films from in situ Raman spectroscopy: they interpreted the formation of neutral TCNQ 0 under electrical field as a proof of the charge transfer. MIM-type CuTCNQ-based memory elements exhibit a bipolar resistive switching [4,7]: by applying a negative bias to the top electrode (TE) typically made of Al with a grounded bottom electrode (BE), the memory element, initially in a high resistance state (HRS or OFF state), switches into a low resistance state (LRS or ON state), i.e. set operation. By reversing voltage polarity, a reset operation is achieved in turning back the structure into HRS state. Many works have been done to uncover the mechanism responsible for this resistance change. Recently, the switching was attributed to the formation and dissolution of Cu filaments through a native oxide layer formed at the interface between CuTCNQ complex and an oxidizable metal (e.g. aluminum) used as top electrode [8–11]. Typical memory elements reported in literature consist in stacks Al(TE)/CuTCNQ/Cu(BE) in cross-point architecture in which the widths of perpendicular top and bottom electrode stripes define the area of the memory structure [7]. In the perspective of high density memory devices, CuTCNQ material has to grow in small interconnect structures with metallic electrodes (e.g. Cu, Al, etc.) compatible to the standard CMOS process (complementary metal-oxide semiconductor). Although CuTCNQ can be prepared by several methods (e.g. reaction of Cu with a TCNQ solution [5,12]; Contents lists available at ScienceDirect journal homepage: www.elsevier.com/locate/jcrysgro Journal of Crystal Growth 0022-0248/$ - see front matter & 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.jcrysgro.2010.08.008 n Corresponding author. Tel.: + 33 491054779; fax: + 33 491054782. E-mail address: christophe.muller@im2np.fr (Ch. Muller). Journal of Crystal Growth 312 (2010) 3267–3275