Characterizations and thermal stability improvement of phase-change memory
device containing Ce-doped GeSbTe
films
Yu-Jen Huang, Min-Chuan Tsai, Chiung-Hsin Wang, Tsung-Eong Hsieh
⁎
Department of Materials Science and Engineering, National Chiao Tung University, 1001 Ta-Hsueh Road, Hsinchu, 30010, Taiwan, ROC
a b s t r a c t
a r t i c l e i n f o
Article history: Received 1 April 2011
Received in revised form 27 November 2011 Accepted 1 December 2011
Available online 11 December 2011 Keywords:
Phase-change memory Germanium antimony telluride Cerium
Doping
Electrical properties
Phase-transition temperature of GeSbTe (GST) chalcogenidefilm was drastically increased from 159 to 236 °C by cerium (Ce) doping (up to 8.6 at.%) without altering the resistivity property of GST. Grain refinement via the solid-solution mechanism and the amplification of p-type semiconducting behavior in Ce-doped GST were ob-served. They were correlated with the enhancement of thermal stability and data retention property of GST as revealed by exothermal and isothermal analyses. Phase-change memory (PCM) device characterized at various temperatures revealed an effective thermal stability improvement on the threshold voltage of PCM device by Ce doping.
© 2011 Elsevier B.V. All rights reserved.
1. Introduction
Phase-change memory (PCM) based on the reversible phase-change property of chalcogenide thinfilms has been recognized as one of the promising candidates for next-generation non-volatile memories due to its low power consumption, high operation speed, high recording density and excellent scalability to nanometer-scale cell sizes. In addition, PCM is also feasible to multi-state memory since the chalcogenide programming layer exhibits a relatively large resistivity difference (about 105times) in between amorphous and crystalline states[1,2]. This is far more significant than the reflectivity change up to 30% for the same phase-change chalcogenides applied to optical memory devices.
Presently, GeSbTe (GST)-based chalcogenides are the most popu-lar programming layer materials for PCM. However, as the device sizes continuously scale down, insufficiency in material properties of GST gradually restraints the device performance, e.g., writing cur-rent, thermal stability, and overwrite capability. Various methods have been proposed to overcome these difficulties and, in addition to adjusting the chemical constitutions of chalcogenides, the modi fi-cation of physical properties by alien-element doping is often employed. Nitrogen (N)[3,4], oxygen (O)[5], silicon oxide (SiOx) [6], silicon (Si)[7], molybdenum (Mo)[8], tin (Sn)[9,10], and bis-muth (Bi)[10] have been added into GST and their feasibilities to PCM have been evaluated. Among these, nonmetallic N element seems to be the most promising one for physical properties
adjustment. As to the metallic dopants, they were reported to acceler-ate the recrystallization racceler-ate of GST[11]. Nevertheless, metallic dop-ants often reduce the resistivity level of amorphous GST and suppress the resistivity ratio of the amorphous and crystalline GST
[8].
Most dopants reported previously[3–7,8–10]are the Group IVA, VA or VIA elements with the electron configurations similar to those of Ge, Sb and Te. In this work, cerium (Ce) is chosen as the dopant of GST for the following reasons: (i) distinctive electron configuration of the Group IIIB rare-earth element; (ii) relatively small electronegativity (χ) in comparison with the elements of GST (χCe= 1.1–1.2;
χGe= 1.8;χSb= 1.9;χTe= 2.1); (iii) relatively large atomic size of Ce
in comparison with Ge, Sb and Te (rCe= 0.185 nm; rGe= 0.125 nm;
rSb= 0.145 nm; rTe= 0.140 nm). In addition to the chemical bond
alter-nation in GST due to the difference inχ for physical property modifica-tion, it is anticipated that Ce may present as the substitutional-type dopant in GST and relevant study would provide a good comparison with the N dopant which has been reported as the interstitial-type solute in GST[4].
2. Experimental details
Electrical properties and microstructures of pristine GST and Ce-doped GST layers are investigated in this work. 150-nm thick thin-film samples were deposited on the thermally oxidized Si substrates by using a sputtering system at background pressure better than 6.7 × 10− 4Pa without substrate heating. Ce doping was realized by the target-attachment method[8]in which the doping concentration was adjusted by the amounts of Ce foils mounted on the GST target
⁎ Corresponding author.
E-mail address:[email protected](T.-E. Hsieh).
0040-6090/$– see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.tsf.2011.12.014
Contents lists available atSciVerse ScienceDirect
Thin Solid Films
(Ge2Sb2Te5; purity: 99.99%; supplier: Tosima, Japan). The Ce foils
(pu-rity: 99.99%) were supplied by Alfa Aesar, USA.
The composition of specimen was analyzed by using an inductive-ly coupled plasma-mass spectrometer (Perkin Elmer SCIEX ELAN 5000) operated at RF power = 1000 W, carrier gas flow rate = 1.3 × 10− 5m3/s, auxiliary gas
flow rate=2×10− 5m3/s and
plasma gasflow rate=2.5×10− 4m3/s. Evolution of microstructures
of the samples were analyzed by an in-house x-ray diffractometer (M18XHF, MAC Science) within Cu-Kαradiation (λ=0.1542 nm) at
scanning rate of 0.033°/s. The microstructures of the samples were also examined by a transmission electron microscope (TEM, Philips Tecnai 20) operating at 200 kV and the energy dispersive spectrome-ter (EDS, Link ISIS 300) attached to TEM was adopted for composition analysis. The method of plan-view TEM sample preparation is briefed as follows. Pristine and Ce-doped GST layers about 50 nm in thickness were deposited on the KBr substrates. After dissolving away the KBr in a dish of de-ionized water, the GST thinfilms were dragged out of the water, mounted on the copper mesh, and immediately trans-ferred to TEM for characterization. Ecopia HMS-3000 Hall-effect ana-lyzer was utilized to analyze the effects of Ce-doping on the charge carrier type and transport properties of GST.
Static-mode current–voltage (I–V) characteristics were determined by using an I–V source meter (Keithley Instruments Inc., Model 2400) equipped with a self-assembly heating apparatus in vacuum ambient with pressure better than 0.13 Pa. Meanwhile, cross-point-type PCM devices with 5-μm contact hole containing pristine GST or Ce-doped GST as the programming layers were prepared. The device structure with the electrical characterization setup is schematically shown in
Fig. 1.
3. Results and discussion
Composition analysis revealed the Ce content in doped GST sam-ples monotonously increases with the increase of area coverage ratio of Ce foils on GST target. For instance, the doped sample pre-pared at the condition of Ce area coverage ratio = 7% resulted in about 8.6 at.% Ce in GST. Similar increasing trend was also observed in the characterization of phase-transition temperature of Ce-doped GST as revealed by subsequent in-situ electrical property measure-ment. Since the increase trend of phase-transition temperature allevi-ated and intermetallic phases seemed to emerge in the samples prepared at area coverage ratio of Ce = 10%, in below we report the analytical results of Ce-doped GST prepared at Ce area coverage ratio = 7% (termed GST7C hereafter). Characterizations for pristine GST sample are also presented for the purpose of comparison.
Fig. 2shows the XRD profiles of as-deposited samples and samples subjected to 300 °C/1-h and 350 °C/1-h annealing treatments. Ab-sence of characteristic peaks in diffraction patterns of as-deposited samples indicates the samples deposited at room temperature are amorphous. High-temperature treatments are known to result in the recrystallization of GST; however,Fig. 2indicates GST transforms from meta-stable face-centered cubic (FCC) to stable hexagonal structure when temperature increases from 300 to 350 °C while the FCC structure is preserved in the GST7C heated up to 350 °C. This il-lustrates Ce doping is able to stabilize the GST in FCC status. Further, a careful inspection of XRD patterns found the Ce doping causes the XRD peaks shift to smaller diffraction angle side. Meanwhile, a calcu-lation of full-width-half maximum (FWHM) of (200)FCCpeaks for the
samples heated at 300 °C for 1 h found that the FWHM of GST7C layers = 1.61° while that of GST = 1.12°. This implies that Ce doping results in the lattice expansion and grain size refinement of GST. The lattice expansion is ascribed to the relatively large atomic radius of Ce in comparison with those of Ge, Sb and Te.
Fig. 3(a) and (b) separately presents the plan-view TEM micro-graphs, both bright- and dark-field images, in conjunction with the selected area electron diffraction (SAED) patterns for GST and GST7C samples subjected to 300 °C/1-h annealing. Grain size reduc-tion in GST7C due to Ce doping can be readily seen, in agreement with the XRD analysis presented previously.
Fig. 3(c) shows the EDX element mapping on a specific area of GST7C sample shown inFig. 3(b). Separation of Ge-rich and SbTe-rich phases can be observed in accord with the element image contrast; however, Ce element mapping indicates a nearly uniform dispersion of Ce in GST7C. Since the XRD analysis shown inFig. 2revealed no Ce-related intermetallic compounds (IMCs) in GST, it implies that Ce pre-sents in a form of solid solution in GST in composition range studied in this work. The Ce atoms should disperse in GST lattice as the substi-tutional solutes due to the comparatively large atomic size of Ce. The stressfield induced by the difference in atomic sizes may inhibit the grain boundary motion and hence result in the grain refinement of GST7C as depicted inFig. 3(b).
Fig. 4(a) depicts the typical resistivity (ρ) and corresponding de-rivative (dρ
dT) profiles for 150-nm thick GST and GST7C layers as a
function of temperature measured at a heating rate of 1 °C per mi-nute. A unique features observed inFig. 4(a) is, unlike other metallic dopants, Ce doping barely affects the levels ofρ for amorphous GST and the resistivity ratio of amorphous and crystalline GST remains the same at about 105 times. At the same temperature level, the
Thermal oxide (150 nm) Si (100)
Bottom electrode (100 nm)
Programming layer (150 nm)
Top electrode (100 nm)
PECVD Si oxide (150 nm)
Source meter (Keithley 2400)
Hot stage
Computer
Temperature
controller
Multimeter
(Fluke 8845)
Thermal oxide (150 nm) Si (100)Hot stage
Fig. 1. Cross-sectional view of PCM device and the electrical characterization setup.
Fig. 2. XRD patterns of as-deposited GST and GST7C layers and the samples subjected to 300 and 350 °C annealing for 1 h.
crystalline GST7C exhibits a higher level ofρ in comparison with the pristine GST. This is ascribed to the grain refinement effect due to Ce doping and will benefit the reduction of writing current when the Ce-doped GST is implanted in the PCM devices.
Since the values ofρ decrease with the increase of temperature as depicted inFig. 4(a), both the pristine and the Ce-doped GSTs, either in amorphous of crystalline state, possess the semiconducting feature. Charge carrier property measurement indicated both samples are p-type semiconductors with the hole concentrations = 3.55 × 1020and
2.92 × 1021cm− 3, respectively. Presumably, the amplification of
acceptor-like characteristics in Ce-doped GST is correlated to the elec-tronic configuration of Ce which is known as [Xe]4f1
5d16s2. The partially-filled inner 4d and 5f orbitals in Ce might provide additional acceptor levels in the bandgap structure of GST for hole carrier genera-tion. The hole concentration increment and the crystallinity improve-ment in crystalline GST7C would benefit the electrical conduction; nevertheless, the presence of grain boundaries and lattice distortion cause by Ce doping discourage the propagation of charge carriers so as to neutralize the conduction improvement. This results in moderate conductivity suppression for crystalline GST7C in comparison with the crystalline GST.
Previous studies reported that the metallic dopants usually sup-press the resistivity level or, promote the conduction of amorphous GST[8–10]. In GST7C sample, the electrons at outermost 6s orbital of Ce likely induce the resistivity drop. However, stressfields caused by the difference of atomic sizes in between Ce and elements in GST matrix would disturb the carrier migration. Such a carrier scattering effect might be a key factor to inhibit the drop ofρ in amorphous GST7C as illustrated inFig. 4(a). We however note this work is a pre-liminary study of Ce-doped GST. Topics such as the effect of dopant's electronic configuration on the electrical properties of doped GST, the effects of Ce doping on bandgap structure and bonding configurations of GST, either in amorphous or crystalline state, remain to be investi-gated in order to characterize the origin of transport property changes in Ce-doped GST.
Fig. 3. Plan-view TEM micrographs of (a) GST and (b) GST7C samples subjected to the annealing at 300 °C for 1 h. Bright-field image (left-hand part), dark-field image (right-hand part) and SAED patterns (upper right-(right-hand side) are presented simultaneously. (c) EDX element mapping for Ge, Sb, Te, and Ce in a specific area of GST7C sample shown in (b). 0 60 120 180 240 300 360 10-3 10-1 101 103 105
dρ
/dT (a
rb
.u
n
its)
R
e
si
stiv
ity
,
(
-cm
)
Temperature(
oC)
Temperature(
oC)
GST GST7C 24 28 32 36 e7 e12 e17 e22 e27 e32 GST GST7CCharacteristic time,
τ (sec)
1/k
BT
188 141 94 47(a)
(b)
ρ
Fig. 4. (a) Typical profiles of ρ and corresponding derivatives (dρ
dT) as a function of
temperature measured at a heating rate of 1 °C per minute for GST and GST7C. (b) A plot of characteristic time versus reciprocal of temperature.
According to the derivative profiles of resistivity depicted inFig. 4(a), recrystallization temperatures (Tx) for GST and GST7C were determined
as 159 and 236 °C, respectively (seeTable 1). This illustrates Ce doping may stabilize the amorphous GST. The increment of Txis ascribed to
the uniformly dispersed Ce solutes in GST matrix that inhibit the grain growth and coalescence during recrystallization.
Presently N doping is widely recognized as the most promising method to modify the physical properties of chalcogenides[3,4]. An-alytical results presented above indicate Ce doping is as effective as N doping to improve the recrystallization property of GST; however, N doping simultaneously increases the resistivity levels of amorphous and crystalline GST[3,4,8]whereas, as depicted byFig. 4(a), Ce doping does not dramatically alter the resistivity levels of amorphous and crys-talline GST. In other words, N doping shifts the resistivity profile toward high-resistivity side while Ce doping shifts the resistivity profile toward the high-temperature side in the plot of resistivity against temperature. When the PCM cells are programmed, adjacent amorphous bits suffer the parasitic heating which might result in the data loss. Such a thermal cross-talk phenomenon has become a concern of PCM under the trend of device scale-down[2]. Thermal resistance is known to correlate with the phonon and electron scattering in the bulk of material. In PCM, ther-mal boundary resistance (TBR) is another issue that considerably affects the temperature distribution of programming layer and the device perfor-mance[12–14]. TBR is resulted from the scattering process at the interface and its influence exaggerates when the difference in resistance across the interface becomes large. According to the resistivity properties presented above, it can be inferred that N doping pronounces the thermal resistance heating and TBR effects in the devices whereas Ce doping might alleviate the thermal cross-talk in between PCM cells in the trend of device scale-down. Further, in a view of fabrication Ce doping can be achieved by con-ventional sputtering using the composite GST target containing appropri-ate amount of Ce. Though N doping can also be realized by simply allowing the N2gasflow during deposition, it nevertheless induces the
re-active sputtering that the etching effect caused by inadequate N2gasflow
would degrade thefilm quality. As a result, Ce is a promising dopant for the modification of GST properties as well as the enhancement of PCM performance.
Both exothermal experiment in conjunction with Kissinger theory
[15]and isothermal experiment in conjunction with Arrhenius for-mulation[16]were carried out in this study and the activation ener-gies, Eaand Eaf, corresponding to these two types of experiments were
calculated accordingly. As summarized inTable 1, Ce doping escalates both the values of Ea and Eaf. During the isothermal process, the
change ofρ is in essential driven by the percolation effect[8]and this process can be elucidated by a characteristic time,τ, in the plot ofρ versus time. A plot of τ versus the reciprocal of T for GST and GST7C is shown inFig. 4(b). This allows us to determine the maxi-mum temperatures of sample retention by extrapolating method
[16]. For 10-year retention, the maximum temperatures for GST and GST7C samples were found to be 76 °C and 170 °C, respectively. In conjunction with the values of Ea and Eaf listed in Table 1, it
illustrates that Ce doping may effectively enhance the thermal stability of GST layers and thereby benefit the data retention of PCM.
Fig. 5(a) and (b) presents the static-mode I–V characteristics of PCM devices containing GST and GST7C programming layers at vari-ous test temperatures. It can be seen that the switching threshold voltage (Vth) increases with the Ce doping (Vth= 1.90 V for GST and
Vth= 3.30 V for GST7C at room temperature). Since the proposal of
chalcogenides for PCM devices, the threshold switching phenomenon has attracted numerous research interests and its origin is still in de-bate. The explanations include microscopic phase change[17], ther-mally induced instabilities [18], carrier generation by impact ionization[19,20], and the mobility gap of an amorphous material
[21,22]. In the viewpoint of phase-change process, Vthis related to
the requirement of a significant current passing through the cell and the generation of heat by Joule effect to induce the phase transition. Presumably, GST7C is able to produce a larger Vthdue to its larger Ea
than that of GST. In accord with the profiles shown inFig. 5(a) and (b), the temperature dependence of Vthcan be obtained and is plotted
inFig. 5(c). Though each device features the decrease in Vthwith the Table 1
A list of the values of Tx, Ea, Eaf, and temperatures for 10-year retention of amorphous
state of GST and GST7C layers.
Samples Tx(°C) Ea(eV) Eaf(eV) Temperatures for 10-year retention
(°C) GST 159 2.05 ± 0.1 2.3 ± 0.3 76 GST7C 236 4.53 ± 0.1 4.63 ± 0.2 170 0 1 2 3 4 5 6 0 20 40 60 80
Current (mA)
0 20 40 60 80Current (mA)
Voltage (V)
Room temp. @ 50oC @ 100oC @ 150oC @ 200oC 0 1 2 3 4 5 6Voltage (V)
-40 0 40 80 120 160 200 240 0.0 0.6 1.2 1.8 2.4 3.0 3.6Threshold Voltage (V)
Temperature (
oC)
GST GST7C(c)
(a)
(b)
Room temp. @ 50oC @ 100oC @ 150oC @ 200oCFig. 5. Static I–V profiles for PCM devices containing (a) GST and (b) GST7C as the programming layers at various test temperature. (c) The temperature dependence of Vthof PCM devices.
increase of temperature, GST7C device is superior to GST device in terms of stability in Vth at various temperatures. Another feature
shown inFig. 5(c) is that the Vthof GST7C device is less sensitive to
the temperature change in comparison with GST device. Without the incorporation of dopants, pristine GST exhibits an obvious change in grain size during the amorphous-to-crystalline transition as illus-trated byFig. 3(a). By forming the obstacle of grain boundary mo-tions, Ce stabilizes the microstructure of GST7C (seeFig. 3(b)) and consequently leads to the stabilization of physical properties for doped sample subjected to temperature change. Analytical results presented above confirm that Ce doping may enhance the thermal stability of amorphous GST and result in a better PCM device stability
[6,7]. Further, Ce doping effectively refines the grains of GST and this would benefit the scale-down of PCM devices and the reduction of writing current as revealed by the resistivity property presented in
Fig. 4(a).
4. Conclusions
In summary, this work demonstrates a metallic dopant type, Ce, which may effectively enhance the thermal stability of GST without degrading its electrical properties. Analytical results indicated that Ce-doping effectively refines the microstructure of GST via the solid-solution mechanism. Enhancement of p-type semiconducting proper-ty was observed in Ce-doped GST, implying the dopant's electronic configuration likely plays a key role in the property modification of GST. The difference in ρ for amorphous and crystalline Ce-doped GST remains at aboutfive orders of magnitude without the loss of phase-change reversibility. At the same temperature level, Ce-doping was found to escalate the level ofρ of crystalline GST which, in turn, benefits the writing current reduction for PCM applications. As compared with pristine GST device, GST7C device exhibited a better stability in switching threshold property. With such distinctive electri-cal properties, Ce-doped GST can be a promising chalcogenide for PCM device applications with better signal contrast preservation and high-density signal storage capability. Thefindings reported in this work may also benefit the development of optical storage media if the analysis on optical properties of Ce-doped GST were accomplished.
Acknowledgments
The work is supported by the National Science Council (NSC), Taiwan, R.O.C., under contract No. NSC97-2221-E-009-029-MY3. The authors would like to thank Dr. Chia-Hung Hsu at NSRRC, Hsinchu, Taiwan, R.O.C., for the supports of XRD analysis. The supports of plasma-enhanced chemical vapor deposition (PECVD) by National Nano Device Laboratories (NDL) and reactive ion etching (RIE) by Gigastorage Co., at Hsinchu, Taiwan, R.O.C., for PCM device preparation and TEM analysis by Materials Analysis Technology Inc. at Chupei, Taiwan, R.O.C., are also deeply acknowledged.
References
[1] S. Lai, T. Lowrey, Tech. Dig., IEDM Tech. Dig., 2001, p. 803.
[2] A. Pirovano, A.L. Lacaita, A. Benvenuti, F. Pellizzer, S. Hudgens, R. Bez, IEDM Tech. Dig. (2003) 699.
[3] Y. Lai, B. Qiao, J. Feng, Y. Ling, L. Lai, Y. Lin, T. Tang, B. Cai, B. Chen, J. Electron. Mater. 34 (2005) 176.
[4] Y.K. Kim, J.H. Baeck, M.-H. Cho, E.J. Jeong, D.-H. Ko, J. Appl. Phys. 100 (2006) 083502. [5] N. Matsuzaki, K. Kurotsuchi, Y. Matsui, O. Tonomura, N. Yamamoto, Y. Fujisaki, N. Kitai, R. Takemura, K. Osada, S. Hanzawa, H. Moriya, T. Iwasaki, T. Kawahara, N. Takaura, M. Terao, M. Matsuoka, M. Moniwa, IEDM Tech. Dig. (2005) 738. [6] S.W. Ryu, J.H. Oh, J.H. Lee, B.J. Choi, S.Y. Hwang, S.K. Hong, C.S. Hwang, H.J. Kim,
Electrochem. Solid-State Lett. 9 (2006) G259.
[7] T. Zhang, Z. Song, F. Rao, G. Feng, B. Liu, S. Feng, B. Chen, Jpn. J. Appl. Phys. 46 (2007) L247.
[8] Y.J. Huang, Y.C. Chen, T.E. Hsieh, J. Appl. Phys. 106 (2009) 034916. [9] W.D. Song, L.P. Shi, X.S. Miao, T.C. Chong, Appl. Phys. Lett. 90 (2007) 091904. [10] T.J. Park, S.Y. Choi, M.J. Kang, Thin Solid Films 515 (2007) 5049.
[11] Liqiu Men, Junji Tominaga, Hiroshi Fuji, Takashi Kikukawa, Nobufumi Atoda, Jpn. J. Appl. Phys. Part 1 40 (2001) 1629.
[12] E. Bozorg-Grayeli, J.P. Reifenberg, K.W. Chang, M. Panzer, K.E. Goodson, Thermal and Thermomechanical Phenomena in Electronic Systems (ITherm), Las Vegas, U.S.A., June 2–5, 12th IEEE Intersociety Conference, 2010, p. 5501263. [13] J.P. Reifenberg, K.W. Chang, M.A. Panzer, S.B. Kim, J.A. Rowlette, M. Asheghi, H.S.
Philip Wong, K.E. Goodson, IEEE Electron Device Lett. 31 (2010) 56.
[14] E.K. Kim, S.I. Kwun, S.M. Lee, H. Seo, J.G. Yoon, Appl. Phys. Lett. 76 (2000) 3864. [15] H.E. Kissinger, Anal. Chem. 29 (1957) 1702.
[16] R.E. Reed-Hill, R. Abbaschian, Physical Metallurgy Principles, 3rd ed., PWS, Boston, 1992, p. 241.
[17] S.R. Ovshinsky, Phys. Rev. Lett. 21 (1968) 1450. [18] A.C. Warren, IEEE Trans. Electron Devices 20 (1973) 123.
[19] D. Adler, M.S. Shur, M. Silver, S.R. Ovshinsky, J. Appl. Phys. 51 (1980) 3289. [20] D. Adler, H.K. Henisch, N. Mott, Rev. Mod. Phys. 50 (1978) 209.
[21] A. Pirovano, A.L. Lacaita, A. Benvenuti, F. Pellizzer, R. Bez, IEEE Trans. Electron Devices 51 (2004) 452.
