Investigation of switching region in superlattice phase change memories T Ohyanagi and N Takaura Citation: AIP Advances 6, 105104 (2016); doi: 10.1063/1.4964729 View online: http://dx.doi.org/10.1063/1.4964729 View Table of Contents: http://aip.scitation.org/toc/adv/6/10 Published by the American Institute of Physics AIP ADVANCES 6, 105104 (2016) Investigation of switching region in superlattice phase change memories T Ohyanagia and N Takaura Hitachi, Ltd Research & Development Group, 1-280, Higashi-koigakubo, Kokubunji-shi, Tokyo 185-8601, Japan (Received 29 March 2016; accepted 29 September 2016; published online October 2016) We investigated superlattice phase change memories (PCMs) to clarify which regions were responsible for switching We observed atomic structures in a superlattice PCM film with a stack of GeTe / Sb2 Te3 layers using atomically resolved EDX maps, and we found an intermixed region with three atom species of the Ge, Sb and Te around the top GeTe layer under the top electrode We also found that a device with a GeTe layer on an Sb2 Te3 layer without superlattice structure had the same switching characteristics as a device with a superlattice PCM, that had the same top GeTe layer We developed and fabricated a modified superlattice PCM that attained ultra low Reset / Set currents under 60 µA © 2016 Author(s) All article content, except where otherwise noted, is licensed under a Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/) [http://dx.doi.org/10.1063/1.4964729] Phase change memories (PCMs) based on Ge, Sb and Te (GeSbTe) are promising candidates for replacing flash memories as next-generation solid-state memories Nanostructured PCMs based on the stacked structure of [GeTe / Sb2 Te3 ]1,2 (superlattice PCMs) have been proposed These memories exhibit much lower Reset / Set currents than conventional PCMs using alloy materials such as Ge2 Sb2 Te5 A number of experimental demonstrations of the superlattice PCMs have yielded positive results, including lower Reset / Set currents,4 superior endurance characteristics,5 and excellent thermal stability.6 However, the switching mechanism is not yet fully understood at the atomic level Some studies have been concluded on the atomic structures of superlattice PCMs using high angle annular dark field scanning transmission microscopy (HAADF-STEM)7 and X-ray diffraction (XRD).8 Some attempts to use computer simulations have been reported,9–13 as well These reports suggested that the switching was due to vertical displacements of Ge atoms at the interface of GeTe layers and Sb2 Te3 layers in the superlattice The driving force of the Ge displacements has been attributed to charge injections,9 electric field,10 thermal activation,11 and stresses in the superlattice films.12,13 However, we need to clarify whether these displacements occurred in all interfaces in the superlattice, in a part of the interfaces, or in other regions In this study, we performed experiments to identify which regions are responsible for the switching in the superlattice PCMs We fabricated several types of superlattice PCMs with different GeTe / Sb2 Te3 combinations using both Ge50Te50 and Ge30Te70 compositions A comparison of the switching characteristics in these different superlattice PCM structures revealed that the region around the top layer of the superlattice was responsible for the switching On the basis of these results, we fabricated a superlattice PCM with an improved structure that exhibited Reset / Set currents lower than 60 µA Figure 1(a) shows a typical superlattice PCM structure with top and bottom electrodes made of tungsten (W) The superlattice PCM was composed of three blocks: a bottom layer, a GeTe / Sb2 Te3 superlattice, and a top layer The bottom layer must be Sb2 Te3, 14 in order to fabricate superlattice PCM films to orient the growth along the direction We set the thickness of the bottom layer to 10 nm.15,16 The GeTe and Sb2 Te3 were prepared by sputtering at 200 ◦ C, and we used aElectronic mail: takasumi.oyanagi.pv@hitachi.com 2158-3226/2016/6(10)/105104/8 6, 105104-1 © Author(s) 2016 105104-2 T Ohyanagi and N Takaura AIP Advances 6, 105104 (2016) FIG (a) Schematic image of a superlattice PCM film with top and bottom electrodes (b) HAADF-STEM image of structure (I) in Table I (c) (i) Atomically resolved EDX maps of structure (I) around the top GeTe layer and the top electrode of W (ii) Atomically resolved EDX maps of structure (I) in the superlattice film 300 mmΦ silicon (Si) wafers to fabricate the superlattice PCMs The block of the superlattice was a stacked structure of GeTe and Sb2 Te3 In our production conditions using 300 mmΦ wafers, the top layer must be GeTe because peeling between the top layer and the top electrode of tungsten (W) would always occur if we used Sb2 Te3 as the top layer We used two types of GeTe sputtering targets, Ge50Te50 and Ge30Te70 We found that Ge50Te50 and Ge30Te70 had a 50 : 50 and a 30 : 70 ratio of Ge and Te in atomic percent, respectively The phase diagram of the GeTe shows that Ge30Te70 was the mixture of Te and GeTe, but the crystallization temperature of the GeTe on the Sb2 Te3 template was reportedly much lower than that without the Sb2 Te3 template.17 We determined that the crystalline structure of the GeTe on the Sb2 Te3 was different from that without the Sb2 Te3 template Moreover, Soeya et al reported that GeTe (Ge50Te50) in the superlattice had vacancies.16 We determined that Ge30Te70 on the Sb2 Te3 had more vacancies than Ge50Te50 on the Sb2 Te3 We already reported that Reset currents of the Gex Te100-x / Sb2 Te3 (x < 50) superlattice PCMs were much lower than those of Ge50Te50 / Sb2 Te3 superlattice PCMs,5 and these were due to easy displacement of the Ge atoms by the many vacancies in the superlattice PCMs The detailed fabrication processes of the superlattice PCM were described in our previous paper.18 We fabricated six types of superlattice PCM structures to clarify the region responsible for switching and to enable low-current switching These six structures are listed in Table I and numbered (I) to (VI) We measured the switching characteristics of the superlattice PCMs using the pulse generator of Agilent 81110A with voltage tolerance within % to the applied voltage and the oscilloscope of Agilent 54853A to read currents We used a rectangular pulse that had a pulse width of 50 ns in both Reset / Set operations 105104-3 T Ohyanagi and N Takaura AIP Advances 6, 105104 (2016) TABLE I Superlattice PCM structures used in this study No superlattice PCM film structure (I) (II) (III) (IV) (V) (VI) Sb2 Te3 = 10 nm / [Ge50Te50 = nm / Sb2 Te3 = nm] × / Ge30Te70 = nm Sb2 Te3 = 10 nm / [Ge30Te70 = nm / Sb2 Te3 = nm] × / Ge30Te70 = nm Sb2 Te3 = 50 nm / Ge30Te70 = nm Sb2 Te3 = 50 nm / Ge50Te50 = nm Sb2 Te3 = 10 nm / [Ge50Te50 = nm / Sb2 Te3 = nm] × / Ge50Te50 = nm Sb2 Te3 = 10 nm / [Ge50Te50 = nm / Sb2 Te3 = nm] × / Ge30Te70 = nm Figure 1(b) shows a HAADF-STEM image of the superlattice PCM film of structure (I) This image was as-deposited one, and the initial condition of the superlattice PCM of structure (I) was the high resistance states (HRS) This image was measured at an accelerating voltage of 200 kV We checked the TEM images were not changed during measurements by the electron beams Figure 1(c) shows an atomically resolved EDX map The upper part (i) of Fig 1(c) corresponds to the image in the region around the top GeTe layer and the top electrode of W, and the lower part (ii) corresponds to the image in the region of the superlattice The (i) and (ii) in Fig 1(c) correspond to the square regions in Fig 1(b) Figure 1(c) shows a mixed map with three atom species of Te, Sb and Ge and maps with a single atom of Te, Sb or Ge The arrow in Fig 1(c)(i) indicates the regions of the top 3-nm-thick GeTe layer From a macroscopic point of view such as that of the XRD,9 the arrangements of the GeTe and Sb2 Te3 in the superlattice PCM film are considered well ordered The peaks of the XRD spectra show that the structure of the Sb2 Te3 was Te-Sb-Te-Sb-Te and that of the GeTe was Ge-Te-Ge-Te or Ge-Te-Te-Ge As shown in Fig 1(c)(ii), one region has an obvious absence of Sb in the superlattice: the GeTe layer The ordered structures are also shown to be kept in the superlattice However, we found an unclear region near the top electrode W in Fig 1(c)(i) In this region, three atom species of Te, Sb and Ge seem to be intermixed We conclude that this region is related to the switching of the superlattice PCM Moreover, the Ge atoms were seen in the W region, but we considered that was not related to the switching of the superlattice PCM Though the top structure of the W and the superlattice of GeTe / Sb2 Te3 was as same as that of the single GeTe and W, the switching characteristics of the GeTe were not different from that of the superlattice PCMs Because the GeTe was widely known as a phase change material,19 but described the above the switching characteristics of the superlattice PCM were not different from those of the conventional phase change materials However, one possibility was arising that diffusions of the GeTe into W were originated from the stress in the superlattice films It was reported that stresses in superlattice films were driving forces of moving of Ge atoms.12,13 However, we considered that to fabricate superlattice PCMs on a 300 mmΦ wafer it was needed to reduce stresses in the superlattice films and W in order to prevent peeling off films Moreover, if there were larger stresses in superlattice films and W, we found more intermixing layers at not only around the top GeTe layer but also into the beneath layers Here, we set up a hypothesis that the region in the superlattice PCM responsible for the switching was around the top GeTe layer First, to investigate the effect of the top layer of the superlattice PCM on the switching, we compared the switching characteristics of structure (II), (III), and (IV) summarized in Table I The device structures are shown in Fig (a) Structure (II) had a superlattice in which 4-nm-thick Sb2 Te3 and 1-nm-thick GeTe were repeated eight times; this is a common superlattice PCM structure.20 In this study, we utilized Ge30Te70 as the GeTe layer in the superlattice Ge30Te70 with a thickness of nm was also used on the top layer to prevent peeling from the top electrode of W Structure (III) had a Ge30Te70 layer nm thick on an Sb2 Te3 layer of 50 nm, and structure (IV) had a Ge50Te50 layer nm thick on an Sb2 Te3 layer of 50 nm instead of the superlattice in structure (II), as shown in Fig 2(a) The thickness of the Sb2 Te3 , 50 nm, was the same as the sum of the superlattice and the bottom layer in structure (II) Figures (b) and (c) show the typical Reset and Set characteristics of device (II), (III), and (IV) as a function of voltage The Reset / Set characteristics of devices (II) and (III) were almost the same, that is, the Reset voltage was around 0.7 V, and the Set voltage was around 0.8 V with very steep switching within 0.1 V Added to that, a change in the resistance over the order of of the structure (II) was almost same as that of the structure (III) We conclude that 105104-4 T Ohyanagi and N Takaura AIP Advances 6, 105104 (2016) FIG (a) Schematic image of the superlattice PCM of structures (II) and (III) (b) Reset characteristics of the superlattice PCM of structures (II) and (III) (c) Set characteristics of the superlattice PCM of structures (II) and (III) the switching region of the superlattice PCM occurred around the top GeTe layer Moreover, the Reset / Set characteristics of structure (IV) were similar to the superlattice switching characteristics with steep switching,16 but the Reset / Set voltages of the structure (IV) were larger than those of the structures (II) and (III) We considered that these differences in the Reset / Set voltage arise from the difference in the top GeTe layers Next, we measured the Reset / Set characteristics of the superlattice PCMs of structures (IV) and (V) The superlattice and the bottom Sb2 Te3 layer of both these structures were the same We prepared the top layer of Ge50Te50 for structure (IV) and Ge30Te70 for structure (V), as shown in Fig 3(a) The typical Reset / Set voltages of structure (V) were much smaller than those of structure (IV), as shown in Figs 3(b) and (c) We conclude that these differences come from the different composition of the top GeTe layer, which supports our hypothesis 105104-5 T Ohyanagi and N Takaura AIP Advances 6, 105104 (2016) FIG (a) Schematic image of the superlattice PCM of structures (IV) and (V) (b) Reset characteristics of the superlattice PCM of structures (IV) and (V) (c) Set characteristics of the superlattice PCM of structures (IV) and (V) However, there is a question that is true only when switching in the nm thick GeTe layer leads to orders of magnitude change in electrical resistance In our device layout using structure (IV) or (V), the stacked film of the W and the superlattice were routed to the electrical wiring and the 105104-6 T Ohyanagi and N Takaura AIP Advances 6, 105104 (2016) electrical pad: the size of the pad was 80 àm ì 80 µm, as shown in Fig 4(a) In Fig 4(a), Pad A is connected to the top electrode, and Pad B is connected to the bottom electrode For the example of structure (II) shown in Fig 2(a), the cross sectional image of line C is the left side of Fig 2(a), and that of line D is Fig 4(b) When the nearest top layer of the GeTe was related to the resistance change, the under layer of the superlattice film was regarded as the bottom electrode Large areas of the superlattice including an 80 àm ì 80 àm pad originated from the resistance change, and we found that this led to the orders of magnitude change FIG (a) Schematic image of the device layout of superlattice PCM (b) Cross sectional image of the superlattice PCM of structure (II), if we cut line D 105104-7 T Ohyanagi and N Takaura AIP Advances 6, 105104 (2016) Considering the aforementioned results, we designed an improved structure to achieve lower switching based on structure (VI) We felt superlattice films did not need to be thick for lower switching because the region in the superlattice PCM responsible for the switching was around the top GeTe layer A thinner superlattice film could make the dry-etching process easier Shintani et al reported that when the superlattice cycle (N) was one using [SnTe / Sb2 Te3 ] × N superlattice PCMs, the Reset power was the smallest up to N=8.21 This report showed that thinner superlattice PCM films resulted in lower Reset power However, because of our fabrication line restrictions, the thickness of the top electrode of W should be over 25 nm To prevent peeling between the superlattice films and the top electrode, the thickness of the superlattice films must be almost exactly the same as the top electrode, around 25 nm We selected [Ge50Te50 = nm / Sb2 Te3 = nm] × as the superlattice film The bottom Sb2 Te3 had a thickness of 10 nm, and the top GeTe had a thickness of nm, the same as the superlattice PCMs described earlier We utilized the Ge30Te70 as the top GeTe layer The minimum unit of Te-Sb-Te-Sb-Te was 1-nm-thick Sb2 Te3 The total thickness of the superlattice PCM of structure (I) was 23 nm, almost the same as that of the 25-nm-thick W of the top electrode, whose device structure is shown in Fig 5(a) Figures 5(b) and (c) show the Reset / Set characteristics of this superlattice PCM The Reset voltage was 0.6 V and the Set voltage was V These are almost the same as those of structure (V) However, the resistance of the low resistance states (LRS) was about 100 kΩ higher than those of structure (V) We conclude that this is because the superlattice PCM films had a similar thickness to the W electrode and that the thinner Sb2 Te3 layer resulted in improved quality of the superlattice films Higher resistance in the LRS reached lower Reset / Set currents Figure 5(d) shows the Reset / Set characteristics of this superlattice PCM indicated in the FIG (a) Schematic image of the superlattice PCM of structure (I) (b) Reset characteristics of the superlattice PCM of structure (I) (c) Set characteristics of the superlattice PCM of structure (I) (d) Reset / Set characteristics of the superlattice PCM of structure (I) shown in the current sweep 105104-8 T Ohyanagi and N Takaura AIP Advances 6, 105104 (2016) current sweep We obtained the best data on ultra-low Reset / Set currents of 60 µA, using rectangular pulses of 50 ns width on both Reset / Set operations However, a detailed switching mechanism of a superlattice PCM was not cleared in this study, but we considered that it was very important to narrow down the switching region of the top GeTe and the Sb2 Te3 In summary, we investigated superlattice PCMs to clarify which regions were responsible for the switching We found an intermixed region around the top GeTe layer on the superlattice PCM film with the superlattice of [GeTe / Sb2 Te3 ] The switching characteristics of the device of the single GeTe on the single Sb2 Te3 were the same as those of the superlattice PCM, indicating that the region in the superlattice PCM responsible for the switching was around the top GeTe layer, rather than the superlattice We proposed a modified superlattice PCM structure with Sb2 Te3 = 10 nm / [Ge50Te50 = nm / Sb2 Te3 = 1nm] × / Ge30Te70 = nm, and ultra-low Reset / Set currents under 60 µA were subsequently attained ACKNOWLEDGMENTS This work was performed as “the Ultra-Low Voltage Device Project” funded and supported by NEDO and METI A part of this work was done by “the Low-power Electronics Association and Project (LEAP).” A part of the device processing was operated by AIST, Japan J Tominaga, P Fons, A V Kolobov, T Shima, T C Chong, R Zhao, H K Lee, and L Shi, Jpn J Appl Phys 47, 5763 (2008) R E Simpson, P Fons, A V Kolobov, T Fukaya, M Krbal, T Yagi, and J Tominaga, Nature Nanotech 6, 501 (2011) Y Fujisaki, Jpn J Appl Phys 49, 100001 (2010) N Takaura, T Ohyanagi, M Tai, M Kinoshita, K Akita, T Morikawa, H Shirakawa, M Araidai, K Shiraishi, Y Saito, and J Tominaga, IEDM Tech Dig 2014, 29.2 T Ohyanagi, N Takaura, M Tai, M Kitamura, M Kinoshita, K Akita, T Morikawa, S Kato, M Araidai, K Kamiya, T Yamamoto, and K Shiraishi, IEDM Tech Dig 2013, 30.5 M Tai, T Ohyanagi, M Kinoshita, T Morikawa, K Akita, S Kato, and N Takaura, VLSI Tech Dig 2014, 22.4 J Momand, R Wang, J E Boschker, M A Verheijen, R Calarco, and B J Kooi, Nanoscale 7, 19136 (2015) T Ohyanagi, M Kitamura, M Araidai, S Kato, N Takaura, and K Shiraishi, Appl Phys Lett 104, 252106 (2014) N Takaura, T Ohyanagi, M Kitamura, M Tai, M Kinoshita, K Akita, T Morikawa, S Kato, M Araidai, K Kamiya, T Yamamoto, and K Shiraishi, Symposium on VLSI Technology T130 (2013) 10 T Egami, K Johguchi, S Yamazaki, and K Takeuchi, Jpn J Appl Phys 53, 04ED02 (2014) 11 D Bang, H Awano, J Tominaga, A V Kolobov, P Fons, Y Saito, K Makino, T Nakano, M Hase, Y Takagaki, A Giussani, R Calarco, and S Murakami, Scientific Report 4, 5727 (2014) 12 X Zhou, J Kalikka, X Ji, L Wu, Z Song, and R E Simpson, Adv Mater 28, 3007 (2016) 13 J Kallika, X Zhou, E Dicher, S Wall, J Lu, and R E Simpson, Nat Commun 7, 11983 (2016) 14 J Tominaga, A V Kolobov, P Fons, X Wang, Y Saito, T Nakano, M Hase, S Murakami, J Herfort, and Y Takagaki, Sci Tech Adv Mater 16, 014402 (2015) 15 private communication with Dr Tominaga of AIST, Japan 16 S Soeya and T Shintani, J Appl Phys 112, 034301 (2012) 17 R E Simpson, P Fons, A V Kolobov, M Krbal, and J Tominaga, Appl Phys Lett 100, 021911 (2012) 18 T Ohyanagi, N Takaura, M Kitamura, M Tai, M Kinoshita, K Akita, M Morikawa, and J Tominaga, Jpn J Appl Phys 52, 05FF01 (2013) 19 S Loubriat, D Muyard, F Fillot, A Roule, M Veillerot, J P Barnes, P Gergaud, L Vandroux, M Verdier, and S Maitrejean, Microelectric Engineering 88, 817 (2011) 20 J Tominaga, R E Simpson, P Fons, and A V Kolobov, Proc of EPCOS 2010, pp.54 (2010) 21 T Shintani, S Soeya, and T Saiki, ECS Transactions 64, 71 (2014) ... was arising that diffusions of the GeTe into W were originated from the stress in the superlattice films It was reported that stresses in superlattice films were driving forces of moving of Ge... down the switching region of the top GeTe and the Sb2 Te3 In summary, we investigated superlattice PCMs to clarify which regions were responsible for the switching We found an intermixed region. .. the superlattice, in a part of the interfaces, or in other regions In this study, we performed experiments to identify which regions are responsible for the switching in the superlattice PCMs