Vol. 34, No. 5
Journal of Semiconductors
May 2013
Reactive ion etching of Si2 Sb2 Te5 in CF4 /Ar plasma for a nonvolatile phase-change memory device Li Juntao(李俊焘)1; 2; , Liu Bo(刘波)1; , Song Zhitang(宋志棠)1 , Yao Dongning(姚栋宁)1 , Feng Gaoming(冯高明)3 , He Aodong(何敖东)1; 2 , Peng Cheng(彭程)1; 2 , and Feng Songlin(封松林)1 1 State Key Laboratory of Functional Materials for Informatics, Shanghai Institute of Microsystem and Information Technology,
Chinese Academy of Sciences, Shanghai 200050, China 2 University of Chinese Academy of Sciences, Beijing 100049, China 3 United Laboratory, Semiconductor Manufacturing International Corporation, Shanghai 201203, China
Abstract: Phase change random access memory (PCRAM) is one of the best candidates for next generation nonvolatile memory, and phase change Si2 Sb2 Te5 material is expected to be a promising material for PCRAM. In the fabrication of phase change random access memories, the etching process is a critical step. In this paper, the etching characteristics of Si2 Sb2 Te5 films were studied with a CF4 /Ar gas mixture using a reactive ion etching system. We observed a monotonic decrease in etch rate with decreasing CF4 concentration, meanwhile, Ar concentration went up and smoother etched surfaces were obtained. It proves that CF4 determines the etch rate while Ar plays an important role in defining the smoothness of the etched surface and sidewall edge acuity. Compared with Ge2 Sb2 Te5 , it is found that Si2 Sb2 Te5 has a greater etch rate. Etching characteristics of Si2 Sb2 Te5 as a function of power and pressure were also studied. The smoothest surfaces and most vertical sidewalls were achieved using a CF4 /Ar gas mixture ratio of 10/40, a background pressure of 40 mTorr, and power of 200 W. Key words: reactive ion etching; phase-change material; Si2 Sb2 Te5 DOI: 10.1088/1674-4926/34/5/056001 PACC: 7360F; 8160 mixture gas. A smooth surface was successfully obtained using the optimization approach described below.
1. Introduction Nowadays, phase change random access memory (PCRAM) has been regarded as one of the most promising non-volatile memories, and has received more and more attention because of its superior performance and other meritsŒ1;2 . It was devised by Ovshinsky in 1968Œ3 based on the rapid reversible phase change effect in some materials under the influence of an electric current pulse, and the different resistances between crystalline and amorphous states define the logic state of an individual bit. Phase change Si2 Sb2 Te5 material, expected as a promising material for PCRAM, possesses a wider band-gap comparing to Ge2 Sb2 Te5 . The band-gap width of amorphous and polycrystalline Si2 Sb2 Te5 are determined to be 0.89 and 0.62 eV by means of Fourier transform infrared spectroscopyŒ4 . The material possesses a low threshold current from amorphous to polycrystalline state in voltage–current measurement, and shows a good data retention. These properties prove Si2 Sb2 Te5 is a potential materialŒ4; 5 . In this paper, the reactive ion etching (RIE) process of Si2 Sb2 Te5 films in CF4 /Ar plasma is described. The etch rate and surface roughness were examined systematically as a function of pressure, power, and Ar concentration in the CF4 /Ar
2. Experiment In this study, Si2 Sb2 Te5 films were deposited with the radio frequency (RF)-magnetron sputtering method using single element targets at room temperature. The thickness of the films was about 400 nm measured by a cross-sectional scanning electron microscope (SEM, Hitachi S-4700). The compositions of films were determined by means of energy dispersive spectroscopy (EDS). Shipley 6809 photo-resist was used for pattern definition. An Oxford 80plus RIE system with a maximum RF power of 600 W was used to etch the Si2 Sb2 Te5 films. The etch gas ratio was controlled by mass flow controllers, and the gas pressure in the chamber was adjusted by a clapper valve. The temperature of the sample holder was controlled by heat transfer fluid (Hexid) and held at 30 ıC. The experimental control parameters were the gas flow rate, chamber background pressure, CF4 /Ar ratio and the incident RF power applied to the lower electrode. A total flow rate of CF4 CAr was 50 sccm throughout the experiment, while the CF4 /Ar ratio was varied as an optimization parameter. Etching depths were measured using a surface profile-
* Project supported by National Key Basic Research Program of China (Nos. 2010CB934300, 2011CBA00607, 2011CB9328004), the National Integrate Circuit Research Program of China (No. 2009ZX02023-003), the National Natural Science Foundation of China (Nos. 60906004, 60906003, 61006087, 61076121, 61176122, 61106001), the Science and Technology Council of Shanghai (Nos. 11DZ2261000, 11QA1407800), and the Chinese Academy of Sciences (No. 20110490761). † Corresponding author. Email:
[email protected];
[email protected] Received 25 August 2012, revised manuscript received 3 December 2012 © 2013 Chinese Institute of Electronics
056001-1
J. Semicond. 2013, 34(5)
Li Juntao et al.
Fig. 1. Etch rate of the Si2 Sb2 Te5 and RMS roughness as a function of CF4 /Ar gas mixture ratio.
meter. The surface morphology and patterning of the mesa structure were examined using SEM, and the surface roughness was examined using atomic force microscopy (AFM).
3. Results and discussion Figure 1 shows the etch rate as a function of the CF4 /Ar gas mixture ratio. The etchings were carried out at a constant pressure of 50 mTorr and an application of 200 W. The etch rate decreased monotonically with decreasing CF4 concentration indicating its importance in defining the material removal rate. In the plasma system, when an energetic electron strikes a neutral gas molecule, it can excite the molecule to a higher energy state. These energetic F containing molecules, known as free radicals, cause most of the chemical etching of Si2 Sb2 Te5 Œ6 . In this experiment, as the content of CF4 went down, the concentration of energetic F decreased, so the etch rate of Si2 Sb2 Te5 decreased in turn. The phenomenon observed is consistent with this mechanism. Compared with Ge2 Sb2 Te5 , we found that the etch rate of Si2 Sb2 Te5 was faster, this should be related with the different boiling points of the etch products, such as SiF4 [boiling point (bp): 65 ıC] and GeF4 [bp: 36:5 ıC]. The lower boiling point of SiF4 makes it easier to be removed from the chamber, as the volatile product of Si2 Sb2 Te5 , this property leads to a faster etch rate of Si2 Sb2 Te5 Œ7 . The quality of etched surfaces is very important for the device fabrication processŒ8 . Many short-circuit defects are due to RIE pillars caused by micro-masks. The smoother the etched surface is the better contact between Si2 Sb2 Te5 and the top/bottom electrode is obtained, which can result in a low contact resistance. The sidewall is also important for the device fabrication process, particularly for the nanoscale etching of Si2 Sb2 Te5 films. Therefore, etched surfaces with a smooth surface, vertical sidewall, and low sidewall roughness are preferred to meet the requirements of the high-density memory devices. Etch residues are not observed on the sidewalls or the film surfaces for all conditions. As the Ar concentration is increased, both the etch slopes and the root-mean-square (RMS) roughness of the etched surfaces shows improvement. When the CF4 /Ar ratio is 40/10, the etched surface is very rough and pillars formed because of micro masks can be observed. As the CF4 /Ar ratio was decreased to 10/40, an almost vertical
etch slope was obtained, and the etched surface showed significantly improved smoothness. In order to validate these results, AFM images of the etched surfaces under different gas mixture are shown in Fig. 2. When the CF4 /Ar ratio is 40/10, the RMS value of this etched surface is 3.38 nm. It reduces to 1.67 nm when the CF4 /Ar ratio increases to 10/40. This phenomenon should be attributed to the increase in Ar concentration. The role of Ar is to remove the non-volatile etch products deposited on the substrate by physical bombardment. It is evident that ion bombardment contributes positively to improving the smoothness in the etching processŒ9 . For the fabrication of PCRAM devices, etch selectivity (the ratio of etch rates) of Si2 Sb2 Te5 films to insulating materials is a key parameter in the etching process. In this experiment, SiO2 films were prepared by plasma enhanced chemical vapour deposition. Etchings were carried out at a constant pressure of 50 mTorr and applying power of 200 W. As shown in Fig. 3, the selectivity decreases along with the Ar concentration which indicates that the concentration of F has a greater impact on Si2 Sb2 Te5 . The etch rate of the Si2 Sb2 Te5 film as a function of power is shown in Fig. 4. The etch rate decreases linearly with RF power. The decreasing etch rate with increasing power may be related to the plasma sheath layer that exists in the chamber and the influence of non-volatile etch products. When the thickness of the plasma sheath increases, the distance crossed by the radicals to reach the substrate also increased. On the other hand, more fluorine radicals lead to the polymer forming species that eventually deposit as unwanted masking materials on the etched surface. Thus, the etch rate appeared to decreaseŒ10 . From Fig. 5, it is found that the surface of higher power etched Si2 Sb2 Te5 films is much smoother than that of lower power etched ones. In lower power conditions, the kinetic energy of Ar radicals is too low to remove the chemical products in time which causes a rough surface. However, this problem is resolved in high power conditions. The etch rate and RMS roughness of the Si2 Sb2 Te5 films as a function of pressure are shown in Fig. 6. The experiments were carried out when the CF4 /Ar ratio was 40/10 and the PF power kept at 200 W. The etch rate increased with chamber pressure and then decreased. The highest etch rate happened under 60 mTorr. The pressure dependence of the etch rate should be dominated by the active abundance of neutral etchant species. In general, the ion energy and the direction of physical bombardment to the specimens are determined by the direct current (DC) bias voltage which is strongly influenced by the chamber pressure. When the pressure is lower than 60 mTorr, the etch rate is mainly dominated by the active abundance of neutral etchant species, the concentration of radicals increases with the gas pressure, which leads to the increase of etch rate. On the other hand, when the gas pressure is higher than 60 mTorr, according to the collisional plasma sheath modelŒ11 , a collisional effect should be considered. When the pressure increases, the mean free path of the charged particles decreases and hence the dc bias will be lower. As a result, the physical bombardment to the substrate by positive ions becomes lower which leads to a decrease of etch rate. The effect of pressure on the surface roughness was also examined using AFM, and the corresponding RMS roughness
056001-2
J. Semicond. 2013, 34(5)
Li Juntao et al.
Fig. 2. AFM images of the etched Si2 Sb2 Te5 surface with a CF4 /Ar ratio of (a) 40/10, (b) 30/20, (c) 20/30, and (d) 10/40.
Fig. 3. Etch selectivity of Si2 Sb2 Te5 to SiO2 as a function of CF4 /Ar ratio.
Fig. 4. Etch rate of the Si2 Sb2 Te5 film as a function of power.
Fig. 5. SEM images of the Si2 Sb2 Te5 surface after etching under different powers. (a) 100 W. (b) 150 W. (c) 200 W. (d) 250 W.
Fig. 6. Etch rate and RMS roughness of the Si2 Sb2 Te5 film as a function of pressure.
056001-3
J. Semicond. 2013, 34(5)
Li Juntao et al.
Fig. 7. AFM images of the etched Si2 Sb2 Te5 surface at different pressures. (a) 30 mTorr. (b) 40 mTorr. (c) 50 mTorr. (d) 60 mTorr. (e) 70 mTorr.
values are shown in Fig. 7. The etched surface is rough under a background pressure of 30 mTorr (RMS roughness measured 6.11 nm), and it becomes smoother as the pressure increases. As stated above, as the DC bias decreases, the physical bombardment by positive ions decreases and enhances the chemical activity. At a pressure of 40 mTorr, the surface is smoothest with an RMS roughness value of 1.00 nm. However, etching at higher pressure (> 40 mTorr) created a rough surface. This is probably due to the re-deposition of etch products. The ion bombarding energy is too low to remove the re-depositions which act as micro-masks resulting in a rough surface Œ12 14 .
4. Conclusion Reactive ion etching of Si2 Sb2 Te5 thin films with a photoresist mask was studied using a CF4 /Ar gas mixture in inductively coupled plasma. The etch rate of Si2 Sb2 Te5 films in a CF4 /Ar plasma decreased with the decrease of CF4 con-
centration at constant background pressure and power. Ar helped to promote the etching process as it removed the non-volatile products by physical bombardment resulting in a smooth surface. Meanwhile, the selectivity of Si2 Sb2 Te5 films to SiO2 decreased. Etched features of Si2 Sb2 Te5 films in CF4 /Ar gas mixture were best when the CF4 /Ar ratio is 10/40, and a smooth surface and a vertical sidewall were obtained. The chamber pressure and power influenced the etch rate and etched surface roughness. A smooth surface and a vertical sidewall were achieved using the following etching parameters: a CF4 /Ar mixing ratio of 10/40, a base pressure of 50 mTorr, and a power of 200 W. Finally, we have demonstrated that reactiveion etching of Si2 Sb2 Te5 in CF4 /Ar plasma shows good etch characteristics and can be used in the fabrication of PCRAM devices based on Si2 Sb2 Te5 .
References [1] Lam C H. Storage class memory. Solid-State and Integrated Cir-
056001-4
J. Semicond. 2013, 34(5)
Li Juntao et al.
cuit Technology, 2010 [2] Annunziata R, Zuliani P, Borghi M, et al. Phase change memory technology for embedded non volatile memory applications for 90 nm and beyond. IEEE International Electron Devices Meeting, Technical Digest, 2009 [3] Kim I S, Cho S L, Im D H, et al. High performance PRAM cell scalable to sub-20nm technology with below 4F2 cell size. Digest of Technical Papers, Symposium on Extendable to DRAM Applications in VLSI Technology, 2010 [4] Kojima R, Okabayashi S, Kashihara T, et al. Nitrogen doping effect on phase change optical disks. Jpn J Appl Phys, 1998, 37: 2098 [5] Liu Y B, Zhang T, Niu X M, et al. Si2 Sb2 Te5 phase change material studied by an atomic force microscope nano-tip. Journal of Semiconductors, 2009, 30(6): 063003 [6] Kojima R, Yamada N. Acceleration of crystallization speed by Sn addition to Ge–Sb–Te phase-change recording material. Jpn J Appl Phys, 2001, 40: 5930 [7] Zhang T, Song Z T, Liu B, et al. Investigation of phase change Si2 Sb2 Te5 material and its application in chalcogenide random
access memory. Solid-State Electron, 2007, 51: 950 [8] Chinoy P B. Reactive ion etching of benzocyclobutene polymer films. IEEE Trans Components, Parking, and Manufacturing Technology, Part C, 1997, 20(3): 99 [9] Abe H, Yoneda M, Fujiwara N. Developments of plasma etching technology for fabricating semiconductor devices. Jpn J Appl Phys, 2008, 47: 1435 [10] Plank N O V, Cheung R. Functionalization of carbon nanotubes for molecular electronics. Microelectron Eng, 2004, 73/74: 578 [11] Feng G M, Liu B, Song Z T, et al. Reactive-ion etching of Ge2 Sb2 Te5 in CF4 /Ar plasma for non-volatile phase-change memories. Microelectron Eng, 2008, 85(8): 1699 [12] Sheridan T E, Goree J. Collisional plasma sheath model. Phys Fluids B, 1991, 3(10): 2796 [13] Knizikevicius R, Kopustinskas V. Influence of temperature on the etching rate of SiO2 in CF4 + O2 plasma. Microelectron Eng, 2006, 83(2): 193 [14] Wolf R, Helbig R. Reactive ion etching of 6H-SiC in SF6 /O2 and CF4 /O2 with N2 additive for device fabrication. J Electrochem Soc, 1996, 143: 1037
056001-5
