High-Resolution Transmission Electron Microscopy Study of

Jan 27, 2011 - Suparna Pal , R. Aggarwal , Vandna Kumari Gupta , Alka Ingale .... Marissa A. Caldwell , Rakesh Gnana David Jeyasingh , H.-S. Philip Wo...
0 downloads 0 Views 4MB Size
LETTER pubs.acs.org/NanoLett

High-Resolution Transmission Electron Microscopy Study of Electrically-Driven Reversible Phase Change in Ge2Sb2Te5 Nanowires Yeonwoong Jung, Sung-Wook Nam, and Ritesh Agarwal* Department of Materials Science and Engineering, University of Pennsylvania, 3231 Walnut Street, Philadelphia, Pennsylvania 19104, United States ABSTRACT: By combining high-resolution transmission electron microscopy (HRTEM) characterization and electrical measurements on a unique device platform, we study the reversible electrically-driven phase-change characteristics of self-assembled Ge2Sb2Te5 nanowires. Detailed HRTEM analyses are used to correlate and understand the effect of full and intermediate structural transformations on the measured electrical properties of the nanowire devices. The study demonstrates that our unique approach has the potential to provide new information regarding the dynamic structural and electrical states of phase-change materials at the nanoscale, which will aid the design of future phase-change memory devices. KEYWORDS: Phase-change nanowire memory, nonvolatile, HRTEM, amorphous, crystalline

hase-change memory (PCM) utilizes the electric field-induced reversible structural change in chalcogenide materials to switch between crystalline (low resistance) and amorphous (high resistance) phases to store information in a rapid and nonvolatile manner.1 Despite extensive investigations of the electrically driven phase-change switching, the underlying mechanisms involved in the relationship between structural and electrical properties in phase-change materials are quite complex and need to be understood more deeply. Despite some success in theoretical approaches2-4 to explain the atomic motions involved at the atomic-scale, visualization of electrically-driven structural transition has been experimentally challenging. Moreover, the direct correlation of structural phases to their electrical states needs to be studied in greater detail. This limitation is mainly because the active phase-change material in thin-film PCM devices are embedded within multiple layers, which prohibits direct probing of structural transformations on a working device with high spatial resolution while under electrical biasing.5 Recently, efforts have been made to characterize the electric field-driven phasechange in stacked thin-film PCM devices mostly based on atomic force microscopy (AFM).6-8 These approaches, however, suffer from two major drawbacks. (1) The techniques require an addition step of transmission electron microscopy (TEM) to investigate the microstructure of materials since AFM is only sensitive to surface topography.7 (2) The devices have to be crosssectioned for TEM characterization, after which they do not operate, therefore making it impossible to study the evolution of structural and electrical properties on the same operating structure.8 In this regard, recently developed self-assembled phasechange nanowire devices9-11 provides intrinsic advantages for direct structural characterization, benefiting from their unique

P

r 2011 American Chemical Society

horizontal planar geometry with no capping layers, which can enable high-resolution TEM (HRTEM) characterization. Here, we present our first efforts towards combined electrical measurement and HRTEM characterization of phase-change Ge2Sb2Te5 nanowire memory devices configured on a TEM grid. Due to the unique device structure, we could correlate the structural phases of phase-change nanowires with electrical properties. We performed HRTEM characterization repeatedly on the same functional phase-change nanowire devices before and after the application of voltage pulses, which revealed the structural evolution driven by electric fields with very high spatial resolution. To enable electrical and structural measurements on the same nanowire device, a specially-designed silicon nitride (SiNx) membrane TEM grid (200 nm thick SiNx film; Ted Pella, Inc.) was used to serve as a platform on which phase-change nanowire memory devices were fabricated. Figure 1 illustrates the device fabrication steps. First, Pt lines (35-40 μm long and ∼1 μm wide) were patterned with even spacing of ∼10 μm on SiNx membrane area (∼200μm  ∼200 μm) by focused ion beam (FIB; FEI DB 235) deposition to be eventually contacted with electrodes connecting to nanowires. In between the Pt lines, narrow trenches (500 nm to 1 μm wide) were fabricated by FIB Gaþ ion milling (Figure 1a). Ge2Sb2Te5 nanowires grown via the vapor-liquid-solid mechanism9 were then mechanically transferred onto the SiNx membrane by rubbing the nanowire-growth substrate against the TEM membrane grid. The dry transfer method is more effective than the solution casting technique Received: December 29, 2010 Published: January 27, 2011 1364

dx.doi.org/10.1021/nl104537c | Nano Lett. 2011, 11, 1364–1368

Nano Letters

LETTER

Figure 1. Fabrication of phase-change nanowire memory devices on SiNx membrane for TEM experiments. (a) SEM image of a SiNx membrane device with trenches fabricated by FIB before the nanowires are transferred. Scale bar, 100 μm. (inset) Magnified SEM image from the region indicated by the red box where trench structures are fabricated in between FIB-deposited Pt electrode lines. Scale bar, 20 μm. (b) SEM image of the phase-change nanowire that crossed the trench after dry transfer from growth substrates. FIB-Pt contact lines are then fabricated to connect the nanowires to the pre-fabricated Pt electrodes. Scale bar, 2.5 μm. (c) TEM image of a notch structure sculpted in the middle of the nanowire crossing the trench for ensuring phase change in this region for detailed TEM characterization. Scale bar, 200 nm. (d) HRTEM image of the nanowire from the notch region showing that single-crystallinity is maintained. Scale bar, 3 nm. (e) Schematic diagram illustrating the TEM observation of the notch structure of the nanowire.

(dropping of nanowires dispersed in a solvent) in transferring clean nanowires as the latter method often leaves behind organic residue and particles inside the trenches. The nanowires that crossed the trenches were electrically contacted with Pt deposition using FIB, and the nanowire-contacted Pt lines were later extended to the supporting region of the grid outside the membrane to enable electrical measurements (Figure 1b).11 Our device geometry offers several important advantages over other reported TEM membrane-based nanodevices.12-14 First, this structure is more suitable for lattice-resolved HRTEM characterization as nanowires crossing the trenches can be directly exposed to electron beam. Second, the trenched device can use relatively thicker SiNx films (200 nm) over the previously reported devices (50 nm),13,14 and are therefore mechanically more stable during the structural and electrical measurements. Recently, Kelvin probe microscopy analysis of phase-change nanowires suggested that the emergence of amorphous domains occur randomly along the length of nanowires as specified by local potential drops.15 This randomness makes it difficult to study the phase-change phenomena especially inside the TEM, which establishes a need to confine the phase-change region into a specific area. With our device design, it is also possible to introduce a local "hot spot" at any specific location along the nanowire suspended across the trench. By reducing the crosssection of the device locally, one can increase the current density and hence Joule heating, which can force the reversible phase change to occur at the hot spot making structural characterization easier inside the TEM.16-18 The local “notch” structure on the nanowire portion suspended across the trench was created by either FIB Gaþ ion milling (10 pA, 30 kV) or electron beam

sculpting (Figure 1c). Despite the FIB-based device fabrication, the single-crystallinity of phase change nanowire is mostly maintained as shown in Figure 1d with few amorphous regions.19 The notch region is monitored by TEM (JEOL 2010F, 200 keV) before and after the repeated application of voltage pulses (ex situ), as illustrated in the schematic in Figure 1e. Prior to the detailed investigation of electrical switching and structural change under applied voltage pulses for RESET (amorphous)/SET(crystalline) transition, we first confirmed the effectiveness of the notch structure for heat localization. When high amplitude voltage pulses were applied, notched nanowires suffered from breakdown as signified by a drastic increase of device resistance to over 100 GΩ (Figure 2a,b). Subsequent SEM characterization typically reveals that only the notched region undergoes mechanical breakdown induced by Joule-heating while the rest region appears intact (Figure 2b). TEM characterization from the notch region after the device breakdown shows that the structure breaks at this region (Figure 2c). This clearly confirms that the smaller-cross section of the notched region serves as a local “hot-spot” that can be used to localize the phase-change region. The multiple steps of electrical measurement and TEM characterization were then carried out with a notched Ge2Sb2Te5 nanowire (thickness ∼180 nm, thinnest part in the notch region, ∼60 nm) to directly correlate the electrical resistivity with the device’s structural state. This initially crystalline nanowire (∼90 kΩ) was electrically switched to a high-resistive RESET state upon the application of voltage pulses (50 ns) with increasing amplitude (Figure 3a). Once the pulse amplitude reaches 2.6 V, the resistance of the nanowire drastically increased to 4.9 MΩ 1365

dx.doi.org/10.1021/nl104537c |Nano Lett. 2011, 11, 1364–1368

Nano Letters

Figure 2. Localized breakdown of a nanowire at the notch region under very high electrical field. (a) Applied voltage pulse versus measured device resistance to show the electrical breakdown of the nanowire after the application of a 9.4 V pulse. Pulse duration is 5 μs. (b) SEM image of the notched nanowire device after structural breakdown. Scale bar, 500 nm. (c) Low-magnification TEM image from the red box in (b) shows that the structural failure. Scale bar, 100 nm.

(Figure 3a, red plot). To ensure the reversibility of the switching phenomena, crystallization voltage pulses (200 ns) were applied to the amorphized device. Upon the application of voltage pulses above 1.4 V, the nanowire switched back to the low-resistive SET state (100 kΩ; Figure 3a, blue plot). The corresponding lowmagnification bright-field TEM images of the notch area at the amorphous RESET state (4.9 MΩ, Figure 3b) displays brighter imaging contrast while the crystalline SET state (100 kΩ, Figure 3c) states displays darker imaging contrast. Such a noticeable difference of the TEM imaging contrast reflects different structural phases and also demonstrates that the notch structure serves as a good hot spot for localizing the phase change area. HRTEM characterization was performed to further reveal the details of the structural phases at the SET and RESET states. In the RESET state (Figure 3d), a completely amorphized region was observed spanning 10-15 nm localized inside the notch along the nanowire length, while the outside thicker regions remain nearly single-crystalline. In the SET state, the notch area was observed to be polycrystalline with coalesced nanoscale grains with different orientations (Figure 3e), explaining the observed low resistance (Figure 3a). The corresponding fast Fourier transforms (FFTs) of the HRTEM images in the insets of Figure 3d,e show diffusive ring patterns (Figure 3d) and randomly oriented diffraction spots (Figure 3e) representing the amorphous and polycrystalline structures, respectively. These experiments confirm that electrical switching followed by the large resistance change is associated with the structural change between crystalline and amorphous phases at the nanoscale. In all TEM characterization, care was taken to avoid structural deformation due to electron beam irradiation.20,21

LETTER

Figure 3. Correlation of electrical resistivity with structural analysis by TEM/HRTEM for a phase change nanowire device. (a) Programming curve obtained from a Ge2Sb2Te5 nanowire device with a notch. (b-e) TEM images obtained from the notch area of the nanowire device with (b) showing the complete amorphous (RESET) state and (c) crystalline (SET) state. Corresponding HRTEM images are shown in (d) RESET and (e) SET with the corresponding fast-Fourier-transform (FFT) diffraction patterns (insets) indicating the amorphous (d) and the polycrystalline (e) structures. Scale bars are (b) 30, (c) 30, (d) 4, and (e) 3 nm.

One area of particular interest is to study the “intermediate” states with resistance in between the two extremities of high (RESET) and low (SET) values. The intermediate structural phases have not been previously characterized, which can be achieved by applying precisely controlled voltage pulses to amorphized or crystallized devices. In this study, we applied voltage pulses to fully amorphized devices to reach intermediate resistance states during recrystallization process and correlate with their electrical properties. Another fully crystallized nanowire device was switched to the fully amorphized state to a resistance of ∼5.3 MΩ. (Figure 4a) The corresponding brightfield HRTEM image of the RESET state shows a uniform bright imaging contrast (Figure 4b,c), again localized in the notch region. The dc I-V sweep from the fully amorphized state (Figure 4a) shows a well-resolved threshold voltage, Vth at ∼1.8 V, following which a linear I-V behavior reappears indicating Joule-heating induced transition to a fully-crystalline phase. The SET state was then switched back to the full RESET state (∼5.2 MΩ), following then applied voltage pulses of increasing amplitude to take the system to a partially recrystallized state (∼700 kΩ). The HRTEM analysis of the partially recrystallized state (Figure 4d) displays an inhomogeneous imaging contrast that clearly shows randomly located nanoscale crystalline grains (∼5-7 nm) dispersed in the amorphous region with lattice 1366

dx.doi.org/10.1021/nl104537c |Nano Lett. 2011, 11, 1364–1368

Nano Letters

LETTER

Figure 4. Characterization of the intermediate states during the switching process. (a) dc I-V sweeps of a nanowire device showing the threshold switching to fully crystalline state starting at different initial states; fully crystalline state with no threshold event (black squares), fully amorphized state with Vth ∼ 1.8 V (blue triangles), and intermediate crystalline state with Vth ∼ 1.2 V (red circles). (b) Low-magnification TEM image of the RESET state. Scale bar, 50 nm. (c) HRTEM image obtained from the region shown in the inset of (b) when the device is in the fully amorphized RESET state (∼5.3 MΩ), and (d) intermediate (partially recrystallized) state (∼700 kΩ). The red box outlines the nanocrystallites embedded in the amorphous matrix. Scale bars for (c) and (d) are 20 nm. (Inset) Lattice-resolved HRTEM image obtained from one of the nanoscale crystallites in (d) showing lattice spacing of 0.30 nm.

fringes (spacing, ∼0.30 nm) corresponding to (002) planes of the metastable FCC structure of Ge2Sb2Te5 (inset of Figure 4d). The observation of randomly oriented small-sized grains supports the nucleation-dominated phase-change mechanism reported in thermally annealed Ge2Sb2Te5 nanowires.22 The dc I-V sweep of the partially recrystallized state (Figure 4a), shows slightly lower initial resistance (∼700 kΩ) followed by threshold switching, Vth, at a lower value (∼1.2 V) than the fully-amorphized state, followed by a transition to the fully recrystalline phase (TEM not shown but similar to Figure 3e). The lowering of the Vth for a partially crystallized device is due to the formation of the nanoscale crystallites in the amorphous matrix, which can produce very high fields in their vicinity leading to lowering of the field required for threshold switching.4 The reversibility of electrically-driven phase-change process was characterized by correlating electrical measurements and HRTEM analysis. Figure 5a-d shows TEM images of the SET and RESET states, which show dark (bright) contrast corresponding to the crystalline (amorphous) state in the notch area. Figure 5e-g shows a series of repeated dc sweeps and programming curves for the same device; the initially crystalline nanowire device with a resistance of 40 kΩ (Figure 5e labeled 1, TEM image in Figure 5a) undergoes a RESET transition through applied voltage pulses (Figure 5f labeled 2, and Figure 5b), and then switched back to the SET state under dc I-V sweep (Figure 5e, labeled 3). The device was switched a few times between the two extreme structural states and the dc I-V sweep of the SET state (Figure 5g, labeled 4) shows that the device is still active. We then fully amorphized the device (HRTEM in Figure 5h) and performed the dc I-V sweep (Figure 5g, labeled 5), where we again observed threshold switching following that

we stopped the sweep when the device reached a resistance level of ∼90 kΩ (Figure 5g), which is slightly higher than the full SET state resistance of ∼40 kΩ. The bright-field TEM (Figure 5c) shows a contrast very similar to the images obtained from the full SET state, but the HRTEM images obtained from the region marked in Figure 5c shows the emergence of large polycrystalline grains in the amorphous matrix (Figure 5i) that are now connected to form a continuous electrical path between the two sides of the crystalline notch structure. This structural state in Figure 5i is quite different from the intermediate state in Figure 4d, which was obtained at a resistance value much higher than the complete SET state with crystalline grains not forming any continuous network. These experiments demonstrate the importance of HRTEM to provide direct correlation between the electrical properties and the intermediate structural phases at the sub-100 nm regime, which is critical to study the underlying mechanism of phase-change. In conclusion, by utilizing the horizontal, open structure of phase-change nanowire devices fabricated on TEM grids with trenches, we have investigated the phase-change process via combined electrical measurement and high spatial resolution TEM characterization. The reversible phase-change process associated with large resistance changes are directly correlated with various structural states of the nanowire devices. Correlation of electrical intermediate states with their high-resolution structure provides direct insights about the observed changes in the threshold voltage and the conduction behavior in a percolated and nonpercolated system. The study shows that our unique approach has the versatility and potential to provide rich information regarding the dynamic structural and electrical states of phase-change materials at the nanoscale, which could be useful for designing future PCMs. 1367

dx.doi.org/10.1021/nl104537c |Nano Lett. 2011, 11, 1364–1368

Nano Letters

Figure 5. (a-d) TEM images corresponding to the nanowire device during repeated switching between SET and RESET states. Scale bars are 50 nm. (e-g) Electrical characterization of the nanowire device. (e) dc I-V sweep from the initial SET state (labeled 1) and in the RESET state (labeled 3) obtained after the application of voltage pulses, shown in (f). (f) RESET transition under applied voltage-pulses (100 ns) (g) dc I-V sweep from the SET state (labeled 1) and after switching it to the RESET state (labeled 2) due to the application of a voltage pulse (4 V, 100 ns). (f) Programming curve obtained from the initial SET state while applying 100 ns pulses with increasing voltage amplitude. (inset) Magnified programming curve just before RESET switching. (g) dc I-V sweep from the SET state following multiple cycles (labeled 4) and after subsequent switching to the RESET state (labeled 5) but stopped before complete crystallization. (h,i) HRTEM images obtained before and after the labeled 5. Scale bars are 5 nm (h) Amorphous phase after RESET transition and (i) emergence of connected polycrystalline regions after stopping the I-V sweep just below complete crystallization, as shown in (g). The regions outlined by dashed lines are amorphous domains while the solid arrow line shows one of the connected crystalline paths.

LETTER

(2) Sun, Z. M.; Zhou, J.; Ahuja, R. Phys. Rev. Lett. 2006, 96, No. 055507. (3) Kolobov, A. V.; Fons, P.; Frenkel, A. I.; Ankudinov, A. L.; Tominaga, J.; Uruga, T. Nat. Mater. 2004, 3, 703. (4) Karpov, I. V.; Mitra, M.; Kau, D.; Spadini, G.; Kryukov, Y. A.; Karpov, V. G. Appl. Phys. Lett. 2008, 92, No. 173501. (5) Raoux, S.; Rettner, C. T.; Chen, Y. C.; Burr, G. W. MRS Bull. 2008, 33, 847. (6) Hamann, H. F.; O’Boyle, M.; Martin, Y. C.; Rooks, M.; Wickramasinghe, K. Nat. Mater. 2006, 5, 383. (7) Kim, H. J.; Choi, S. K.; Kang, S. H.; Oh, K. H. App. Phys. Lett. 2007, 90, No. 083103. (8) Yoon, S. M.; Choi, K. J.; Lee, N. Y.; Lee, S. Y.; Park, Y. S.; Yu, B. G. Jpn. J. Appl. Phys. 2007, 46, L99. (9) Jung, Y.; Lee, S. H.; Ko, D. K.; Agarwal, R. J. Am. Chem. Soc. 2006, 128, 14026. (10) Lee, S. H.; Ko, D. K.; Jung, Y.; Agarwal, R. Appl. Phys. Lett. 2006, 89, No. 223116. (11) Lee, S. H.; Jung, Y.; Agarwal, R. Nat. Nanotechnol. 2007, 2, 626. (12) Meister, S.; Schoen, D. T.; Topinka, M. A.; Minor, A. M.; Cui, Y. Nano Lett. 2008, 8, 4562. (13) Fischbein, M. D.; Drndic, M. Appl. Phys. Lett. 2006, 88, No. 036116. (14) Johnston, D. E.; Strachan, D. R.; Johnson, A. T. C. Nano Lett. 2007, 7, 2774. (15) Yu, D.; Brittman, S.; Lee, J. S.; Falk, A. L.; Park, H. Nano Lett. 2008, 8, 3429. (16) Lacaita, A. L. Solid-State Electron. 2006, 50, 24. (17) Small, E.; Sadeghipour, S. M.; Pileggi, L.; Asheghi, M. 11th Intersoc. Conf. Therm. Thermomech. Phenom. Electron. Syst. (ITHERM) 2008, 28, 1046. (18) Nam, S. W.; Lee, D.; Kwon, M. H.; Kang, D. M.; Kim, C.; Lee, T. Y.; Heo, S.; Park, Y. W.; Lim, K.; Lee, H. S.; Wi, J. S.; Yi, K. W.; Khang, Y.; Kim, K. B. Electrochem. Solid-State Lett. 2009, 12, H155. (19) Tham, D.; Nam, C. Y.; Fischer, J. E. Adv. Mater. 2006, 18, 290. (20) Kooi, B. J.; Groot, W. M. G.; De Hosson, J. T. M. J. Appl. Phys. 2004, 95, 924. (21) Friedrich, I.; Weidenhof, V.; Lenk, S.; Wuttig, M. Thin Solid Films 2001, 389, 239. (22) Lee, S. H.; Jung, Y. W.; Agarwal, R. Nano Lett. 2008, 8, 3303.

’ AUTHOR INFORMATION Corresponding Author

*E-mail: [email protected].

’ ACKNOWLEDGMENT This work was supported by ONR (Grant N000140910116), Materials Structures and Devices Center at MIT, NSF (DMR0706381 and DMR-1002164), and Penn-MRSEC (DMR0520020). Electron microscopy experiments were performed at the Penn Regional Nanotechnology Facility at the University of Pennsylvania. ’ REFERENCES (1) Wuttig, M.; Yamada, N. Nat. Mater. 2007, 6, 824. 1368

dx.doi.org/10.1021/nl104537c |Nano Lett. 2011, 11, 1364–1368