Metal Organic Chemical Vapor Deposition of Phase Change

Feb 24, 2012 - Within the XPS detectivity limits, we observed no evidence of nitrogen .... Indeed, it is known that phase change speed increases and a...
0 downloads 0 Views 3MB Size
Letter pubs.acs.org/NanoLett

Metal Organic Chemical Vapor Deposition of Phase Change Ge1Sb2Te4 Nanowires Massimo Longo,*,† Roberto Fallica,† Claudia Wiemer,† Olivier Salicio,† Marco Fanciulli,†,‡ Enzo Rotunno,§ and Laura Lazzarini§ †

Laboratorio MDM, IMM-CNR, Unità di Agrate Brianza, Via C. Olivetti 2, 20864 Agrate Brianza, (MB), Italy Dipartimento di Scienza dei Materiali, University of Milano Bicocca, Via R. Cozzi, 53, 20126 Milano, Italy § IMEM-CNR, Parco Area delle Scienze 37/A - 43124 Parma, Italy ‡

ABSTRACT: The self-assembly of Ge1Sb2Te4 nanowires (NWs) for phase change memories application was achieved by metal organic chemical vapor deposition, catalyzed by Au nanoislands in a narrow range of temperatures and deposition pressures. In the optimized conditions of 400 °C, 50 mbar, the NWs are Ge1Sb2Te4 single hexagonal crystals. Phase change memory switching was reversibly induced by nanosecond current pulses through metal-contacted NWs with threshold voltage of about 1.35 V. KEYWORDS: Ge1Sb2Te4 nanowires, MOCVD, VLS, phase-change memory obtained on TiAlN at 250 °C and ∼10 mbar under high supersaturation conditions.21 In this work we succeeded in growing self-assembled Ge1Sb2Te4 NWs on SiO2 by MOCVD and demonstrated their switching property in view of their potential application as phase change structures for scaled memory devices. The self-assembled NW growth was performed by a thermal MOCVD AIX 200/4 reactor, exploiting the VLS mechanism induced by Au metal-catalyst nanoislands. The substrates were 4 inch (100) silicon wafers on which a 50 nm thick thermal SiO2 layer is present. The substrates were preliminarily loaded into a thermal evaporator, where an Au film with nominal thickness of 2 nm was deposited by e-beam evaporation, according to the procedures of ref 20. As a result of the Au evaporation, Au nanoislands were formed with size distribution centered around 20 nm. The SiO2 layer was chosen because it prevents the Ge−Sb−Te from growing on substrate areas where the Au nanoislands are not present. The substrates were afterward loaded into the MOCVD growth chamber, employing purified nitrogen as process gas. Electronic grade t etrakisdim et hylaminogermanium ([N(CH 3 ) 2 ] 4 Ge, TDMAGe), trisdimethylaminoantimony ([N(CH 3 ) 2 ]3 Sb, TDMASb) and diisopropyltelluride ((C3H7)3Te, DiPTe) were used as Ge, Sb and Te precursors, respectively, due to their comparable vapor pressures. Reactant partial pressure in the vapor phase was 1.5 × 10−2 mbar for TDMAGe, 4 × 10−3 mbar for TDMASb, 6.5 × 10−2 mbar for DiPTe; total gas flow was 4.2 L/min and deposition time = 10/20 min. Different deposition runs were performed in the temperature range of (300−445)°C and pressure range of (10−300) mbar. Morphological analysis was based on scanning electron microscopy (SEM) observations, performed by a Zeiss Supra

P

hase change memories (PCMs) are the emerging nonvolatile storage devices based on chalcogenide materials in which the programming function depends on the different electrical resistance offered by their amorphous and crystalline phases, reversibly induced by suitable current pulses. In particular, the Ge2Sb2Te5 (GST-225) and Ge1Sb2Te4 (GST124) alloys exhibit suitable crystallization rates and stability at the operating temperatures of PCMs; moreover, their electrical resistance varies by several orders of magnitude and the phase transition is reversible with a high cycling endurance.1−3 A very attractive option involves the use of chalcogenide nanowires (NWs) for highly scaled PCM devices and multilevel memory applications,4 since lowering active material volumes to be programmed requires shorter and less intense current pulses and implies higher cell density. To this aim, single crystal GST-225,5−8 Ge−Sb,9 Sb2Te3,10 GeTe11 and core−shell GST225/GeTe,12 Sb2Te3/GeTe13 phase change NWs have been deposited by thermal evaporation method under vapor− liquid−solid (VLS) mechanism.14 Probably due to the difficulty in varying the Ge/Sb/Te content by evaporation of the elemental species, no NWs other than crystallized in the GST225 phase have been reported; yet, at least as thin film, GST124 demonstrated a higher dynamical range for multistate recording applications.15 Notably, the metal organic chemical vapor deposition (MOCVD) technology offers new possibilities in terms of compositional control, industrial scalability on 12 inch substrates and relatively high deposition rates, so that it is now employed even in scaled PCM applications.16,17 In particular, the use of MOCVD for the self-assembly of NWs for micro- and optoelectronic applications is widely employed,18 notwithstanding the complexity of the involved chemical−physical processes. In the case of amorphous substrates (SiO2 or Si3N4), the VLS mechanism has led to the MOCVD synthesis of crystalline ZnS and CdS NWs.19 As regards to MOCVD-grown chalcogenide nanostructures, our group reported the growth of GeTe NWs by VLS/vapor− quasisolid−solid mechanisms,20 whereas InSbTe NWs were © 2012 American Chemical Society

Received: December 6, 2011 Revised: February 10, 2012 Published: February 24, 2012 1509

dx.doi.org/10.1021/nl204301h | Nano Lett. 2012, 12, 1509−1515

Nano Letters

Letter

Figure 1. (a−e) SEM tilted views showing the effect of deposition temperature (P = 50 mbar). NWs are obtained in the 375−420 °C range. (f) Effect of deposition temperature on chemical composition of the nanostructures (blue squares = Te, red circles = Sb, black triangles = Ge; lines connecting data are a guide for the eye).

Figure 2. (a−e) SEM tilted views showing the effect of deposition pressure (T = 400 °C). NWs are obtained at 50 mbar. (f) Effect of deposition pressure on chemical composition (blue squares = Te, red circles = Sb, black triangles = Ge; lines connecting data are a guide for the eye).

Electrical characterization of single NWs required the deposition of Pt electrodes on both NW ends, employing focused ion beam (FIB); the Ga+ ion FIB source was operated according to the following settings: 30 kV acceleration voltage, 30 pA (fine), or 300 pA (coarse) aperture. An additional Pt strip was deposited in a similar way across two pads and its resistance was measured to rule it out from the whole device resistance. The NW phase change was observed through pulsed I/V (current/voltage) measurements performed by an Agilent Pulse Generator 81110 to supply the voltage pulses, while the current was measured by a shunt resistor and an oscilloscope; the low-voltage resistance of the device was determined by using an Agilent B1517A Source Monitor Unit. A systematic study for the MOCVD synthesis of the nanostructures was performed as a function of the deposition temperature and pressure, whose main results are summarized in Figures 1 and 2. Both temperature and pressure parameters

40 and a JEOL 6400F field emission microscopes at an accelerating voltage of 15 kV. X-ray diffraction (XRD) experiments, both in grazing incidence and Bragg−Brentano configurations, were carried out by means of an ItalStructures HRD3000 diffractometer; the experimental XRD curves were analyzed by a best fit procedure based on the Rietveld method.22 Compositional analysis was performed by total reflection X-ray fluorescence (TXRF) spectra. Analytical and conventional transmission electron microscopy (TEM) studies were performed in a high-resolution, HR, (0.18 nm) field emission JEOL 2200FS microscope, equipped with in-column Ω energy filter, 2 high-angle annular dark-field (HAADF) detectors and X-ray microanalysis (EDS). The NWs were removed from the deposited samples and then dispersed on holey carbon grids for the observation. 1510

dx.doi.org/10.1021/nl204301h | Nano Lett. 2012, 12, 1509−1515

Nano Letters

Letter

Figure 3. (a) SEM cross section showing the NWs deposited at 400 °C, 50 mbar. The Au droplets on top of the NWs are clearly visible; (b,c): statistical distribution of length and diameter for the NWs of sample in (a), where the red line is the best Gaussian data fit; (d) XRD analysis of the sample in (a): the expected spectra from the different GST stable phases (blue for GST-124, red for GST-147, purple for GST-225) and of Au (in green) are also shown.

temperature of 400 °C and pressure of 50 mbar appeared to be the optimal parameters in the explored ranges. Actually, NWs could even be obtained at T = 350 °C and P = 300 mbar by varying the DiPTe/TDMAGe and TDMAGe/TDMASb concentration ratios. In these cases, the density and length of the NWs are sensibly reduced and/or different crystalline phases appeared. Therefore, these conditions were discarded. Anyway, the growth temperature of 400 °C remains one of the lowest reported for the self-assembly of Ge−Sb−Te NWs.4,5 The SEM cross section image of Figure 3 shows the optimized NWs (400 °C, 50 mbar), randomly tilted with respect to the substrate surface. Because of the amorphous character of SiO2, no preferential growth alignment is expected. Many wires present droplets of different size on the top, most likely Au seeds, as expected by the VLS or vapor−solid−solid23 mechanisms, driving the wire formation. An average density of 3.2 × 106 NWs/mm2 was calculated. Ge−Sb−Te plates and grains are present as growth byproduct. The statistical analysis, performed on several SEM observations, gave the result that the NWs are featured by a variable diameter and length. The length distribution is centered on 350 nm (Figure 3b), but NWs as long as 1 μm can be found for a deposition time of 10 min. By also considering the NWs width distribution of Figure 3c, two types of NWs could be identified: (A) NWs featured by mean cross size >50 nm and aspect ratio (AR = length/mean cross size) distribution centered around 5 and (B) NWs featured by mean cross size