- Email: [email protected]
The ultrafast phase-change memory with high-thermal stability based on SiC-doped antimony
Scripta Materialia 129 (2017) 56–60
Contents lists available at ScienceDirect
Scripta Materialia journal homepage: www.elsevier.com/locate/scriptamat
Regular article
The ultrafast phase-change memory with high-thermal stability based on SiC-doped antimony Tianqi Guo a,b,⁎, Sannian Song a,⁎⁎, Le Li a,b, Xinglong Ji a,b, Chang Li a,b, Chang Xu b,c, Lanlan Shen a,b, Yuan Xue a,b, Bo Liu a, Zhitang Song a, Ming Qi a, Songlin Feng a,c a b c
State Key Laboratory of Functional Materials for Informatics, Shanghai Institute of Microsystem and Information Technology, Chinese Academy of Sciences, Shanghai 200050, China University of the Chinese Academy of Sciences, Beijing 100049, China Shanghai Advanced Research Institute, Chinese Academy of Sciences, Shanghai 201210, China
a r t i c l e
i n f o
Article history: Received 28 August 2016 Received in revised form 26 October 2016 Accepted 26 October 2016 Available online xxxx Keywords: High speed Good stability SiC Nanocomposite Phase transformation
a b s t r a c t The contradictory between switching speed and thermal stability has long been a huge challenge for phasechange memory applications. For providing a feasible solution, the rapid-phase-transition material — pure antimony, was incorporated with silicon carbide to complement their advantages. The characterization results elucidate that dopants situated in the grain boundary inhibit the crystal growth and enhance the stability of the temperature-sensitive amorphous state. Convincingly for devices, an ultrafast speed of 7 ns, an operational voltage requirement of only 1.0 V, a high endurance of more than 200 K and a long data retention are all demonstrated to be realizable and repeatable. © 2016 Acta Materialia Inc. Elsevier Ltd. All rights reserved.
1. Introduction Along with the coming era of the Internet of Things and Big Data, Phase-Change Memory (PCM) is one of the best candidates for new class memory that is capable to store the exponential growth data. Being the most matured device of the emerging memory solutions, it combines the characteristics of high density due to excellent scalability, good reliability attributed to long life time, and high speed owing to fast crystallization mechanism [1–3]. Nevertheless, PCM is still placed on the midst of DRAM and NAND for its deficiencies on operation speed and power consumption compared to DRAM [4]. Therefore, material improvements and new cell designs have been devoted many efforts to achieve complete replacement in recent years. As is well known, antimony is a rapid-phase-transition material with a growth-dominated crystallization mechanism. Besides, this kind of Te-free phase change material is the environmental friendly material compared with other alloys containing Te. However, it has high crystallization rates but low archival life stability [5], which is very sensitive to the factors, such as deposition rate and time, residual gas pressure,
⁎ Correspondence to: T. Guo, State Key Laboratory of Functional Materials for Informatics, Shanghai Institute of Microsystem and Information Technology, Chinese Academy of Sciences, Shanghai 200050, China. ⁎⁎ Corresponding author. E-mail addresses: [email protected] (T. Guo), [email protected] (S. Song).
http://dx.doi.org/10.1016/j.scriptamat.2016.10.034 1359-6462/© 2016 Acta Materialia Inc. Elsevier Ltd. All rights reserved.
substrate temperature, etc. [6] Thus the initially deposited antimony thin film is usually the crystalline state as seen in most relevant literatures [7,8]. In view of the fact that SiC doping was previously used to enhance the thermal stability and boost the operation speed of Sb3Te material [9]. Here it was proposed and expected to be an effective dopant to improve the high temperature stability of this single element matter.
2. Experimental Different groups with the increased SiC content were prepared by co-sputtering method using Sb and SiC targets. The concentration of SiC in Sb was varied by changing the radio frequency sputtering power and evaluated by energy dispersive spectroscopy (EDS). Here the low, medium and high doping concentration denote the materials of (SiC)3.0Sb94, (SiC)4.5Sb91 and (SiC)6.0Sb88, respectively in the following sections. The resistance as a function of temperature was in situ measured in a homemade chamber. Then the microstructure was investigated by transmission electron microscopy (TEM) and the bonding situation was explored by X-ray photoelectron spectroscopy (XPS). Finally, T-shaped PCM cells were prepared for device testing and operating demonstration, of which have a bottom-up construction of 190 nm bottom electrodes array and 60 nm phase-change layer, as well as 10 nm TiN and 300 nm Al, serving as the adherent layer and top electrode, respectively.
T. Guo et al. / Scripta Materialia 129 (2017) 56–60
3. Results and discussion Fig. 1(a) shows the resistance evolution as a function of temperature for the deposited films. The resistance-temperature (R-T) curves exhibit a precipitous drop at 196.8 °C, 235.7 °C and 250.6 °C for (SiC)3.0Sb94 (LD), (SiC)4.5Sb91 (MD) and (SiC)6.0Sb88 (HD), respectively. It precisely mirrors the phase transformation from the amorphous state to the hexagonal crystalline phase. In addition, the value of high resistance state (HRS) becomes larger simultaneously with a higher doping level, which is helpful to improve the amorphous thermal stability. Generally, the pure Sb film is low resistance state (LRS) of crystallization under the same treatment conditions, while light SiC doping significantly improves the situation. More noticeably, the amorphous-state Sb films were prepared for convenient comparison by controlling deposition time at the same growth rate as shown in Fig. 1(c), indicating ultrafast amorphous-crystalline phase transition occurs at a relatively low temperature. And then the activation energy of crystallization (Ea) and the temperature for 10-year data retention (T10-yr) can be calculated according to the Arrhenius equation, Ea t ¼ τ exp kB T where t is the time to failure, τ is a proportional time constant and kB is Boltzmann's constant. Here in Fig. 1(b), the failure time is defined as the time when the resistance decreases to a half of its initial value at the specific temperature. It is observed that the T10-yr of low SiC-doped Sb is 89.6 °C approaching that of Ge2Sb2Te5 [10]. And the lifetime goes up with a higher doping concentration rooting from larger activation energy. Additionally, the calculated Ea values are in good agreement with the
57
data determined by Kissinger's method [11],
ln
dT=dt T 2c
! ¼Cþ
Ea kB T c
where dT/dt is the heating rate, Tc is the crystallization temperature and C is a constant. There is a clear increasing trend on the absolute value of the fitting lines' slope in Fig. 1(d), i.e. Ea. So it also provides a visual contrast to the doped films on activation energy. This convincingly illustrates the argument that SiC improves the lifetime of data by enhancing the stability of pure antimony with a larger energy barrier. The TEM bright-field images with corresponding high-resolution TEM (HRTEM) and the associated selected area electron diffraction (SAED) patterns for two different doping levels are presented in Fig. 2. All the samples were annealed at the temperature of 300 °C for 3 min. The obvious decrease on the grain size can be clearly seen from Figs. 2(a) and (d). And in Figs. 2(b) and (e), the medium SiC-doped Sb exhibits about 20 nm-scale grain size, while high SiC-doped is mostly less than 10 nm. Also from both HRTEM images, we can see that the additional Si or C atoms that situated in the grain boundary lead to the increase of local disorder degree. It has the maximum possibility to cause a striking reduction of grain size. Like Si [12], C [13] or N [14] doping, the disorder phase remarkably inhibits the grain growth and supposedly changes the crystallization mechanism through extending the incubation time of the critical nucleus's formation. Meanwhile, the larger fraction of grain boundaries in SiC-doped Sb is the main reason why the film with higher doping concentration has larger resistance that arises from the scattering effect of defects. Further in Figs. 2(c) and (f), the restrain tendency of grain growth can also be confirmed by SAED patterns. And the diffraction rings of two samples can be identified as hexagonal
Fig. 1. (a) The resistance versus temperature measurements for as-deposited SiC-doped Sb films. The heating rate was 20 °C/min. (b) The corresponding Arrhenius extrapolation plots of failure time versus 1/kBT. (c) Temperature dependence of the resistance measured with various heating rates. (d) Kissinger plots for Ea calculation with regard to pure Sb and doped films.
58
T. Guo et al. / Scripta Materialia 129 (2017) 56–60
Fig. 2. TEM BF images with corresponding HRTEM and SAED patterns for two different doping levels: (a)–(c) medium SiC-doped Sb; (d)–(f) high SiC-doped Sb.
antimony phase (PDF #35-0732). Compared with the unsharper patterns and discontinuous diffraction rings in Fig. 2(c), the decreased crystal grain size is inferred from the continuous diffraction rings with an obvious contrast in Fig. 2(f). In terms of device properties, the stability of LRS (SET state) is enhanced by the uniform distribution of fine crystalline grains, which are conducive to a stable reversible control under the electrical operation. At the same time, the thermotolerant character of HRS (RESET state) gets promoted for the improvement of the data retention behavior at high temperatures. Consequently, the above observations and comparisons imply that SiC doping is an effective method to hinder the growth process of pure antimony. And the crystallization regions feature a significant decrease with a higher doping level, thereby inducing much more stable operation process as shown below. The bond environment of doped films was then investigated by Xray photoelectron spectroscopy. As shown in Fig. 3(a), there is almost no peak shift as the doping concentration increases. We can assume that the element of Sb does not have any electronic interaction with C or Si atoms. And the slight offset can be attributed to the error range, which is only 0.04 eV. However, the deposited components maintain a good consistency, from which the same peak position after etching 30 s and 100 s, respectively. In order to get more details about doping elements, the splitting peaks of C 1s are fitted as displayed in short dot lines in Fig. 3(b). Three chemical states can be found with binding energies of 283.27 eV, 284.73 eV and 286.56 eV, respectively. Peak 2 that has maximum height derives from C\\C bonding, and Peak 3 with minimum height originates from (CH2)n bonding [15]. It is worth noting that Peak 1 has the narrowest full width at half maximum (FWHM) and roots in C\\Si bonds [15]. Similar trends are found in films with the high doping level. It powerfully illustrates the existence states of amorphous disorder phase are not only homogeneous C\\C or Si\\Si bonds, but also the heterogeneous C\\Si bonds with strong binding force. These various types of bonds contribute to locking the amorphous state antimony thus enhancing the thermal stability of this temperature-sensitive phase change material. Fig. 4(a) presents the resistance-voltage (R-V) characteristics for test cells based on medium SiC-doped Sb. Obviously, the Set voltage (Uset) and Reset voltage (Ureset) in Figs. 4(a) and (c) have the same trend of slight increase with the decreasing pulse width. Besides, the cells can still readily achieve complete Set and Reset operation at 7 ns pulse width, well satisfying the desired performance of high-speed operation. Particularly it has the extremely low Uset and Ureset under the ultrafast
Fig. 3. XPS spectra of (a) Sb 4d with two doping levels films; (b) C 1s with medium SiCdoped films. And in (a), the dashed and solid lines represent the curves after etching 30 s and 100 s, respectively.
T. Guo et al. / Scripta Materialia 129 (2017) 56–60
59
Fig. 4. The insert in (a) is the cross-sectional diagram of a T-shaped PCM device. Cell resistances measured as a function of the voltage pulse with different pulse widths for (a) (SiC)4.5Sb91based and (c) (SiC)6.0Sb88-based PCM cells. Endurance characteristics of PCM test cells based on (b) medium SiC-doped, (d) high SiC-doped antimony films.
pulse width of 7 ns, which are 0.7 V and 1.0 V, respectively. Moreover, the Reset voltages for different pulse widths are all much lower than those of the conventional Ge2Sb2Te5-based cells [16] and many other memory units based on novel phase-change materials. Also it has not escaped our attention that the resistivity change occupies nearly two orders of magnitude, ensuring a reliable and adjustable logical partition. Then the detailed R-V characteristics of high SiC-doped test cells are displayed in Fig. 4(c). Though both operation speed and switching voltage do not get promoted, a better uniformity of curves under different pulse widths can be achieved with the help of the increased SiC content. Figs. 4(b) and (d) display the endurance performance of the cells using two different contents of silicon carbide. The Reset/Set resistance ratio is about 30 and 60, respectively. They are both applicable enough for logic distinction. For medium doping, the endurance capability keeps about 2.5 × 105 cycles before a set stuck failure. Meanwhile, it should be noted that the Reset resistances during cyclic operation are less stable than Set resistances before the failure. This little fluctuation with about one half magnitude is largely due to the instability of antimony, which is much sensitive to high-temperature operations. Yet in spite of the fact that the cell can only switch reversibly and stably for 8.4 × 104 cycles with the high doping level, the situation gets improved and exhibits a smaller floating range. Because stability and reliability are usually achieved at the expense of operation speed [17], and besides smaller grains possess better uniformity owing to higher doping contents. So from the above experiments, it at least proves that the speed and operational voltage are always occupying huge advantages whether doping level is high or not for SiC-based Sb test cells. 4. Conclusions In summary, SiC-doped Sb alloy was proposed and explored for PCM applications. The combined electrical testing results and microscopic
investigations suggest it may be the one of the excellent phase change materials reported in the past. Since experimental data have put in evidence that adding SiC impurities to pure Sb raises the crystallization temperature and the corresponding activation energy. Thus it further leads to the favorable data retention, especially maintains the ultrafast switching speed and holds a set of low operation voltages in devices. Additionally of greater importance, both fast speed and high stability can be achieved at the same time. The small grain size and much nucleation centers make the main contribution, which is ascribed to the disorder phase that segregates at grain-boundaries. Therefore, SiC doping is a feasible solution to overcome the contradictory nature of traditional phase-change materials, yet regulating and controlling a more optimal component to obtain a higher stability and persistence in devices is going to be the focus of further research.
Acknowledgments This work was supported by the “Strategic Priority Research Program” of the Chinese Academy of Sciences (XDA09020402), National Key Basic Research Program of China (2013CBA01900), National Integrate Circuit Research Program of China (2009ZX02023-003), National Natural Science Foundation of China (61261160500, 61376006, 51201178), and Science and Technology Council of Shanghai (13DZ2295700, 13ZR1447200, 14ZR1447500, 14DZ2294900).
References [1] G.W. Burr, M.J. BrightSky, A. Sebastian, H.-Y. Cheng, J.-Y. Wu, S. Kim, N.E. Sosa, N. Papandreou, H.-L. Lung, H. Pozidis, E. Eleftheriou, C.H. Lam, IEEE J. Emerg. Sel. Top. Circuits Syst. 1–17 (2016). [2] S. Raoux, D. Ielmini, Chem. Rev. 110 (2010) 240–267. [3] M. Wuttig, S. Raoux, Zeitschrift Für Anorg. Und Allg. Chemie 638 (2012) 2455–2465.
60
T. Guo et al. / Scripta Materialia 129 (2017) 56–60
[4] H.-S.P. Wong, S. Raoux, S. Kim, J. Liang, J.P. Reifenberg, B. Rajendran, M. Asheghi, K.E. Goodson, Proc. IEEE 98 (2010) 2201–2227. [5] L. Van Pieterson, M.H.R. Lankhorst, M. Van Schijndel, a.E.T. Kuiper, J.H.J. Roosen, J. Appl. Phys. 97 (2005). [6] N. Kaiser, Thin Solid Films 116 (1984) 259–265. [7] Y. Hu, X. Zhu, H. Zou, J. Zhang, L. Yuan, J. Xue, Y. Sui, W. Wu, S. Song, Z. Song, Appl. Phys. Lett. 108 (2016) 223103. [8] X. Zhu, Y. Hu, H. Zou, J. Zhang, Y. Sun, W. Wu, L. Yuan, L. Zhai, S. Song, Z. Song, Scr. Mater. 121 (2016) 66–69. [9] T. Guo, S. Song, L. Li, L. Shen, B. Wang, B. Liu, Z. Song, M. Qi, S. Feng, Mater. Lett. 169 (2016) 203–206. [10] W. Zhou, L. Wu, X. Zhou, F. Rao, Z. Song, D. Yao, W. Yin, S. Song, B. Liu, B. Qian, S. Feng, Appl. Phys. Lett. 105 (2014) 243113. [11] H.E. Kissinger, Anal. Chem. 29 (1957) 1702–1706.
[12] Y. Gu, Y. Cheng, S. Song, T. Zhang, Z. Song, X. Liu, X. Du, B. Liu, S. Feng, Scr. Mater. 65 (2011) 622–625. [13] X. Zhou, M. Xia, F. Rao, L. Wu, X. Li, Z. Song, S. Feng, H. Sun, ACS Appl. Mater. Interfaces 6 (2014) 14207–14214. [14] X. Zhou, L. Wu, Z. Song, Y. Cheng, F. Rao, K. Ren, S. Song, B. Liu, S. Feng, Acta Mater. 61 (2013) 7324–7333. [15] J. Chastain, R.C. King, J.F. Moulder, Handbook of X-ray Photoelectron Spectroscopy: A Reference Book of Standard Spectra for Identification and Interpretation of XPS Data, Physical Electronics Eden Prairie, MN, 1995. [16] M. Zhu, L. Wu, F. Rao, Z. Song, K. Ren, X. Ji, S. Song, D. Yao, S. Feng, Appl. Phys. Lett. 104 (2014) 053119. [17] W. Wang, D. Loke, L. Shi, R. Zhao, H. Yang, L.-T. Law, L.-T. Ng, K.-G. Lim, Y.-C. Yeo, T.C. Chong, A.L. Lacaita, Sci. Rep. 2 (2012) 360.
Contents lists available at ScienceDirect
Scripta Materialia journal homepage: www.elsevier.com/locate/scriptamat
Regular article
The ultrafast phase-change memory with high-thermal stability based on SiC-doped antimony Tianqi Guo a,b,⁎, Sannian Song a,⁎⁎, Le Li a,b, Xinglong Ji a,b, Chang Li a,b, Chang Xu b,c, Lanlan Shen a,b, Yuan Xue a,b, Bo Liu a, Zhitang Song a, Ming Qi a, Songlin Feng a,c a b c
State Key Laboratory of Functional Materials for Informatics, Shanghai Institute of Microsystem and Information Technology, Chinese Academy of Sciences, Shanghai 200050, China University of the Chinese Academy of Sciences, Beijing 100049, China Shanghai Advanced Research Institute, Chinese Academy of Sciences, Shanghai 201210, China
a r t i c l e
i n f o
Article history: Received 28 August 2016 Received in revised form 26 October 2016 Accepted 26 October 2016 Available online xxxx Keywords: High speed Good stability SiC Nanocomposite Phase transformation
a b s t r a c t The contradictory between switching speed and thermal stability has long been a huge challenge for phasechange memory applications. For providing a feasible solution, the rapid-phase-transition material — pure antimony, was incorporated with silicon carbide to complement their advantages. The characterization results elucidate that dopants situated in the grain boundary inhibit the crystal growth and enhance the stability of the temperature-sensitive amorphous state. Convincingly for devices, an ultrafast speed of 7 ns, an operational voltage requirement of only 1.0 V, a high endurance of more than 200 K and a long data retention are all demonstrated to be realizable and repeatable. © 2016 Acta Materialia Inc. Elsevier Ltd. All rights reserved.
1. Introduction Along with the coming era of the Internet of Things and Big Data, Phase-Change Memory (PCM) is one of the best candidates for new class memory that is capable to store the exponential growth data. Being the most matured device of the emerging memory solutions, it combines the characteristics of high density due to excellent scalability, good reliability attributed to long life time, and high speed owing to fast crystallization mechanism [1–3]. Nevertheless, PCM is still placed on the midst of DRAM and NAND for its deficiencies on operation speed and power consumption compared to DRAM [4]. Therefore, material improvements and new cell designs have been devoted many efforts to achieve complete replacement in recent years. As is well known, antimony is a rapid-phase-transition material with a growth-dominated crystallization mechanism. Besides, this kind of Te-free phase change material is the environmental friendly material compared with other alloys containing Te. However, it has high crystallization rates but low archival life stability [5], which is very sensitive to the factors, such as deposition rate and time, residual gas pressure,
⁎ Correspondence to: T. Guo, State Key Laboratory of Functional Materials for Informatics, Shanghai Institute of Microsystem and Information Technology, Chinese Academy of Sciences, Shanghai 200050, China. ⁎⁎ Corresponding author. E-mail addresses: [email protected] (T. Guo), [email protected] (S. Song).
http://dx.doi.org/10.1016/j.scriptamat.2016.10.034 1359-6462/© 2016 Acta Materialia Inc. Elsevier Ltd. All rights reserved.
substrate temperature, etc. [6] Thus the initially deposited antimony thin film is usually the crystalline state as seen in most relevant literatures [7,8]. In view of the fact that SiC doping was previously used to enhance the thermal stability and boost the operation speed of Sb3Te material [9]. Here it was proposed and expected to be an effective dopant to improve the high temperature stability of this single element matter.
2. Experimental Different groups with the increased SiC content were prepared by co-sputtering method using Sb and SiC targets. The concentration of SiC in Sb was varied by changing the radio frequency sputtering power and evaluated by energy dispersive spectroscopy (EDS). Here the low, medium and high doping concentration denote the materials of (SiC)3.0Sb94, (SiC)4.5Sb91 and (SiC)6.0Sb88, respectively in the following sections. The resistance as a function of temperature was in situ measured in a homemade chamber. Then the microstructure was investigated by transmission electron microscopy (TEM) and the bonding situation was explored by X-ray photoelectron spectroscopy (XPS). Finally, T-shaped PCM cells were prepared for device testing and operating demonstration, of which have a bottom-up construction of 190 nm bottom electrodes array and 60 nm phase-change layer, as well as 10 nm TiN and 300 nm Al, serving as the adherent layer and top electrode, respectively.
T. Guo et al. / Scripta Materialia 129 (2017) 56–60
3. Results and discussion Fig. 1(a) shows the resistance evolution as a function of temperature for the deposited films. The resistance-temperature (R-T) curves exhibit a precipitous drop at 196.8 °C, 235.7 °C and 250.6 °C for (SiC)3.0Sb94 (LD), (SiC)4.5Sb91 (MD) and (SiC)6.0Sb88 (HD), respectively. It precisely mirrors the phase transformation from the amorphous state to the hexagonal crystalline phase. In addition, the value of high resistance state (HRS) becomes larger simultaneously with a higher doping level, which is helpful to improve the amorphous thermal stability. Generally, the pure Sb film is low resistance state (LRS) of crystallization under the same treatment conditions, while light SiC doping significantly improves the situation. More noticeably, the amorphous-state Sb films were prepared for convenient comparison by controlling deposition time at the same growth rate as shown in Fig. 1(c), indicating ultrafast amorphous-crystalline phase transition occurs at a relatively low temperature. And then the activation energy of crystallization (Ea) and the temperature for 10-year data retention (T10-yr) can be calculated according to the Arrhenius equation, Ea t ¼ τ exp kB T where t is the time to failure, τ is a proportional time constant and kB is Boltzmann's constant. Here in Fig. 1(b), the failure time is defined as the time when the resistance decreases to a half of its initial value at the specific temperature. It is observed that the T10-yr of low SiC-doped Sb is 89.6 °C approaching that of Ge2Sb2Te5 [10]. And the lifetime goes up with a higher doping concentration rooting from larger activation energy. Additionally, the calculated Ea values are in good agreement with the
57
data determined by Kissinger's method [11],
ln
dT=dt T 2c
! ¼Cþ
Ea kB T c
where dT/dt is the heating rate, Tc is the crystallization temperature and C is a constant. There is a clear increasing trend on the absolute value of the fitting lines' slope in Fig. 1(d), i.e. Ea. So it also provides a visual contrast to the doped films on activation energy. This convincingly illustrates the argument that SiC improves the lifetime of data by enhancing the stability of pure antimony with a larger energy barrier. The TEM bright-field images with corresponding high-resolution TEM (HRTEM) and the associated selected area electron diffraction (SAED) patterns for two different doping levels are presented in Fig. 2. All the samples were annealed at the temperature of 300 °C for 3 min. The obvious decrease on the grain size can be clearly seen from Figs. 2(a) and (d). And in Figs. 2(b) and (e), the medium SiC-doped Sb exhibits about 20 nm-scale grain size, while high SiC-doped is mostly less than 10 nm. Also from both HRTEM images, we can see that the additional Si or C atoms that situated in the grain boundary lead to the increase of local disorder degree. It has the maximum possibility to cause a striking reduction of grain size. Like Si [12], C [13] or N [14] doping, the disorder phase remarkably inhibits the grain growth and supposedly changes the crystallization mechanism through extending the incubation time of the critical nucleus's formation. Meanwhile, the larger fraction of grain boundaries in SiC-doped Sb is the main reason why the film with higher doping concentration has larger resistance that arises from the scattering effect of defects. Further in Figs. 2(c) and (f), the restrain tendency of grain growth can also be confirmed by SAED patterns. And the diffraction rings of two samples can be identified as hexagonal
Fig. 1. (a) The resistance versus temperature measurements for as-deposited SiC-doped Sb films. The heating rate was 20 °C/min. (b) The corresponding Arrhenius extrapolation plots of failure time versus 1/kBT. (c) Temperature dependence of the resistance measured with various heating rates. (d) Kissinger plots for Ea calculation with regard to pure Sb and doped films.
58
T. Guo et al. / Scripta Materialia 129 (2017) 56–60
Fig. 2. TEM BF images with corresponding HRTEM and SAED patterns for two different doping levels: (a)–(c) medium SiC-doped Sb; (d)–(f) high SiC-doped Sb.
antimony phase (PDF #35-0732). Compared with the unsharper patterns and discontinuous diffraction rings in Fig. 2(c), the decreased crystal grain size is inferred from the continuous diffraction rings with an obvious contrast in Fig. 2(f). In terms of device properties, the stability of LRS (SET state) is enhanced by the uniform distribution of fine crystalline grains, which are conducive to a stable reversible control under the electrical operation. At the same time, the thermotolerant character of HRS (RESET state) gets promoted for the improvement of the data retention behavior at high temperatures. Consequently, the above observations and comparisons imply that SiC doping is an effective method to hinder the growth process of pure antimony. And the crystallization regions feature a significant decrease with a higher doping level, thereby inducing much more stable operation process as shown below. The bond environment of doped films was then investigated by Xray photoelectron spectroscopy. As shown in Fig. 3(a), there is almost no peak shift as the doping concentration increases. We can assume that the element of Sb does not have any electronic interaction with C or Si atoms. And the slight offset can be attributed to the error range, which is only 0.04 eV. However, the deposited components maintain a good consistency, from which the same peak position after etching 30 s and 100 s, respectively. In order to get more details about doping elements, the splitting peaks of C 1s are fitted as displayed in short dot lines in Fig. 3(b). Three chemical states can be found with binding energies of 283.27 eV, 284.73 eV and 286.56 eV, respectively. Peak 2 that has maximum height derives from C\\C bonding, and Peak 3 with minimum height originates from (CH2)n bonding [15]. It is worth noting that Peak 1 has the narrowest full width at half maximum (FWHM) and roots in C\\Si bonds [15]. Similar trends are found in films with the high doping level. It powerfully illustrates the existence states of amorphous disorder phase are not only homogeneous C\\C or Si\\Si bonds, but also the heterogeneous C\\Si bonds with strong binding force. These various types of bonds contribute to locking the amorphous state antimony thus enhancing the thermal stability of this temperature-sensitive phase change material. Fig. 4(a) presents the resistance-voltage (R-V) characteristics for test cells based on medium SiC-doped Sb. Obviously, the Set voltage (Uset) and Reset voltage (Ureset) in Figs. 4(a) and (c) have the same trend of slight increase with the decreasing pulse width. Besides, the cells can still readily achieve complete Set and Reset operation at 7 ns pulse width, well satisfying the desired performance of high-speed operation. Particularly it has the extremely low Uset and Ureset under the ultrafast
Fig. 3. XPS spectra of (a) Sb 4d with two doping levels films; (b) C 1s with medium SiCdoped films. And in (a), the dashed and solid lines represent the curves after etching 30 s and 100 s, respectively.
T. Guo et al. / Scripta Materialia 129 (2017) 56–60
59
Fig. 4. The insert in (a) is the cross-sectional diagram of a T-shaped PCM device. Cell resistances measured as a function of the voltage pulse with different pulse widths for (a) (SiC)4.5Sb91based and (c) (SiC)6.0Sb88-based PCM cells. Endurance characteristics of PCM test cells based on (b) medium SiC-doped, (d) high SiC-doped antimony films.
pulse width of 7 ns, which are 0.7 V and 1.0 V, respectively. Moreover, the Reset voltages for different pulse widths are all much lower than those of the conventional Ge2Sb2Te5-based cells [16] and many other memory units based on novel phase-change materials. Also it has not escaped our attention that the resistivity change occupies nearly two orders of magnitude, ensuring a reliable and adjustable logical partition. Then the detailed R-V characteristics of high SiC-doped test cells are displayed in Fig. 4(c). Though both operation speed and switching voltage do not get promoted, a better uniformity of curves under different pulse widths can be achieved with the help of the increased SiC content. Figs. 4(b) and (d) display the endurance performance of the cells using two different contents of silicon carbide. The Reset/Set resistance ratio is about 30 and 60, respectively. They are both applicable enough for logic distinction. For medium doping, the endurance capability keeps about 2.5 × 105 cycles before a set stuck failure. Meanwhile, it should be noted that the Reset resistances during cyclic operation are less stable than Set resistances before the failure. This little fluctuation with about one half magnitude is largely due to the instability of antimony, which is much sensitive to high-temperature operations. Yet in spite of the fact that the cell can only switch reversibly and stably for 8.4 × 104 cycles with the high doping level, the situation gets improved and exhibits a smaller floating range. Because stability and reliability are usually achieved at the expense of operation speed [17], and besides smaller grains possess better uniformity owing to higher doping contents. So from the above experiments, it at least proves that the speed and operational voltage are always occupying huge advantages whether doping level is high or not for SiC-based Sb test cells. 4. Conclusions In summary, SiC-doped Sb alloy was proposed and explored for PCM applications. The combined electrical testing results and microscopic
investigations suggest it may be the one of the excellent phase change materials reported in the past. Since experimental data have put in evidence that adding SiC impurities to pure Sb raises the crystallization temperature and the corresponding activation energy. Thus it further leads to the favorable data retention, especially maintains the ultrafast switching speed and holds a set of low operation voltages in devices. Additionally of greater importance, both fast speed and high stability can be achieved at the same time. The small grain size and much nucleation centers make the main contribution, which is ascribed to the disorder phase that segregates at grain-boundaries. Therefore, SiC doping is a feasible solution to overcome the contradictory nature of traditional phase-change materials, yet regulating and controlling a more optimal component to obtain a higher stability and persistence in devices is going to be the focus of further research.
Acknowledgments This work was supported by the “Strategic Priority Research Program” of the Chinese Academy of Sciences (XDA09020402), National Key Basic Research Program of China (2013CBA01900), National Integrate Circuit Research Program of China (2009ZX02023-003), National Natural Science Foundation of China (61261160500, 61376006, 51201178), and Science and Technology Council of Shanghai (13DZ2295700, 13ZR1447200, 14ZR1447500, 14DZ2294900).
References [1] G.W. Burr, M.J. BrightSky, A. Sebastian, H.-Y. Cheng, J.-Y. Wu, S. Kim, N.E. Sosa, N. Papandreou, H.-L. Lung, H. Pozidis, E. Eleftheriou, C.H. Lam, IEEE J. Emerg. Sel. Top. Circuits Syst. 1–17 (2016). [2] S. Raoux, D. Ielmini, Chem. Rev. 110 (2010) 240–267. [3] M. Wuttig, S. Raoux, Zeitschrift Für Anorg. Und Allg. Chemie 638 (2012) 2455–2465.
60
T. Guo et al. / Scripta Materialia 129 (2017) 56–60
[4] H.-S.P. Wong, S. Raoux, S. Kim, J. Liang, J.P. Reifenberg, B. Rajendran, M. Asheghi, K.E. Goodson, Proc. IEEE 98 (2010) 2201–2227. [5] L. Van Pieterson, M.H.R. Lankhorst, M. Van Schijndel, a.E.T. Kuiper, J.H.J. Roosen, J. Appl. Phys. 97 (2005). [6] N. Kaiser, Thin Solid Films 116 (1984) 259–265. [7] Y. Hu, X. Zhu, H. Zou, J. Zhang, L. Yuan, J. Xue, Y. Sui, W. Wu, S. Song, Z. Song, Appl. Phys. Lett. 108 (2016) 223103. [8] X. Zhu, Y. Hu, H. Zou, J. Zhang, Y. Sun, W. Wu, L. Yuan, L. Zhai, S. Song, Z. Song, Scr. Mater. 121 (2016) 66–69. [9] T. Guo, S. Song, L. Li, L. Shen, B. Wang, B. Liu, Z. Song, M. Qi, S. Feng, Mater. Lett. 169 (2016) 203–206. [10] W. Zhou, L. Wu, X. Zhou, F. Rao, Z. Song, D. Yao, W. Yin, S. Song, B. Liu, B. Qian, S. Feng, Appl. Phys. Lett. 105 (2014) 243113. [11] H.E. Kissinger, Anal. Chem. 29 (1957) 1702–1706.
[12] Y. Gu, Y. Cheng, S. Song, T. Zhang, Z. Song, X. Liu, X. Du, B. Liu, S. Feng, Scr. Mater. 65 (2011) 622–625. [13] X. Zhou, M. Xia, F. Rao, L. Wu, X. Li, Z. Song, S. Feng, H. Sun, ACS Appl. Mater. Interfaces 6 (2014) 14207–14214. [14] X. Zhou, L. Wu, Z. Song, Y. Cheng, F. Rao, K. Ren, S. Song, B. Liu, S. Feng, Acta Mater. 61 (2013) 7324–7333. [15] J. Chastain, R.C. King, J.F. Moulder, Handbook of X-ray Photoelectron Spectroscopy: A Reference Book of Standard Spectra for Identification and Interpretation of XPS Data, Physical Electronics Eden Prairie, MN, 1995. [16] M. Zhu, L. Wu, F. Rao, Z. Song, K. Ren, X. Ji, S. Song, D. Yao, S. Feng, Appl. Phys. Lett. 104 (2014) 053119. [17] W. Wang, D. Loke, L. Shi, R. Zhao, H. Yang, L.-T. Law, L.-T. Ng, K.-G. Lim, Y.-C. Yeo, T.C. Chong, A.L. Lacaita, Sci. Rep. 2 (2012) 360.