Sebastien Loubriat, Emanuel Gourvest, Giovanni Betti Beneventi, Helene Feldis,Sylvain Maitrejean, Sandrine Lhostis, Anne Roule, Olga Cueto, Gilles Reimbold,
Ludovic Poupinet, Thierry Billon, Barbara De Salvo, Daniel Bensahel,Pascale Mazoyer, Roberto Annunziata, Paola Zuliani, and Fabien Boulanger
Abstract—In this letter, we present a study on the electricalbehavior of phase-change memories (PCMs) based on a GeTeactive material. GeTe PCMs show, first, extremely rapid SEToperation (yielding a gain of more than one decade in energy perbit with respect to standard GST PCMs), second, robust cycling,up to 1 × 105, with 30-ns SET and RESET stress time, and third, abetter retention behavior at high temperature with respect to GSTPCMs. These results, obtained on single cells, suggest GeTe as apromising alternative material to standard GST to improve PCMperformance and reliability.
Index Terms—Chalcogenide, germanium, GeTe, nonvolatilememories, phase-change (PC) memories (PCMs), telluride.
I. INTRODUCTION
PHASE-CHANGE (PC) memory (PCM) devices are oneof the most promising solutions for the replacement of
standard floating-gate devices [1]. The main advantages arelow voltages, fast read/write, good scalability, and low cost.The active material which is commonly used for PCM is theGe2Sb2Te5 (GST) [1], [2]. For consumer applications, data-retention performance has to be guaranteed for ten years at85 ◦C, and GST complies with this request. On the otherhand, still, open questions remain on how to ensure betterdata-retention performances and even address, with PCM, theembedded memory market. In order to fulfill the request of
Manuscript received January 20, 2010; revised February 10, 2010. Date ofpublication April 5, 2010; date of current version April 23, 2010. The review ofthis letter was arranged by Editor T. Wang.
Color versions of one or more of the figures in this letter are available onlineat http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/LED.2010.2044136
125 ◦C fail temperature after ten years, alternative PC materialsare required. In the literature, a huge number of different PCmaterials have been investigated [3], [4]. In a previous work [5],we have shown that the GeTe material, on full-sheet deposition,presents higher crystallization temperature (i.e., 185 ◦C) thanGST (i.e., 145 ◦C). In [6], tests on a static tester on full-sheetGeTe evidenced a very fast crystallizing process, showing aminimum 30-ns time for the stoichiometric composition. In[7], tests on devices confirmed this very fast SET operation,down to 1-ns stress time. However, all papers lack informationon reliability. In this letter, we compare results obtained onPCM cells based on GeTe and GST, with the same pillarcell architecture. In particular, SET and RESET operations,endurance, and data retention are assessed.
II. EXPERIMENTAL RESULTS AND DISCUSSION
A. Amorphous Phase Stability in Fullsheet Depositions
In order to evaluate the stability of the amorphous phase,100-nm-thick cosputtered amorphous GST and stoichiometricGeTe (53 : 47 ratio by RBS measurements) thin films weredeposited on Si/SiO2 substrates.
Note that the PC materials have been deposited with a dcmagnetron sputtering tool from monotargets of GST and GeTe,respectively. These films have been processed under argonatmosphere with a pressure of 0.005 mbar and a cathode powerof 100 W at room temperature. The deposition rates for GSTand GeTe are 6 and 6.2 Å/s, respectively. The samples forthe resistivity measurements in Fig. 1 were under isothermalcondition while the resistivity was constantly measured. Theexperiment was repeated at different bake temperatures, andthe results are shown in Fig. 1(a) and (b) for GST and GeTe,respectively. It appears that GeTe retains the amorphous stateat higher temperature compared with GST.
By extrapolation from these measurements, based on theArrhenius law, it follows that GST provides a maximum failtemperature after ten years of ∼75 ◦C, with an activationenergy of Ea = 3.13 eV, while the maximum fail temperatureof GeTe is ∼105 ◦C, with Ea = 3.2 eV. These results rep-resent an upper bound of the intrinsic retention properties ofPC materials when integrated in actual devices [8]–[10]. In-deed, as-deposited materials are perfectly amorphous, while the
PERNIOLA et al.: ELECTRICAL BEHAVIOR OF PHASE-CHANGE MEMORY CELLS BASED ON GeTe 489
Fig. 1. Resistivity measurements on full-sheet depositions of (a) GST and(b) GeTe. Measurements are made with a four-probe equipment and a Keithley4200 parametric analyzer. In the same timescale, different temperatures arescreened for GST and GeTe. GeTe confirms superior amorphous phase stability.
melt-quenched amorphous material in PCM can have crys-talline seeds (formed during the quench process) and is sur-rounded by the crystalline matrix: the crystallization processin actual devices can be facilitated. Note in Fig. 1 the differentshapes of isothermal measurements: in GeTe, as soon as a crys-talline nucleus is formed, crystal growth is almost instantaneousand allows a contrast amorphous/crystalline resistivity of morethan four decades; in GST, the crystallization process is muchslower and around two decades of resistivity drop is visible atthe transition from the amorphous phase.
B. PCM Single Cells Performance and Reliability
GeTe and GST have been integrated in a simple pillar devicearchitecture (see inset of Fig. 2) with a 300-nm-wide W pillarin direct contact with a 100-nm-thick PC layer. A 20-nmin situ-deposited TiN layer and the upper top electrode finallycomplete the cell stack. A 200 ◦C thermal annealing in N2
environment is performed at fab out to establish the device inthe SET state, before starting the electrical characterization.
1) Program Characteristics: The program characteristics ofthe integrated test structures were measured using a dedicatedhigh-frequency setup as described in [5]. By using a pulse gen-erator and an active probe, it was possible to read, on a 100-Ωload resistor, the cell current for pulses down to 30 ns (2-nsrise/fall times).
The relative programming speed of GST- and GeTe-baseddevices is shown in Fig. 2. The SET pulse length was increasedfrom a minimum of 30 ns to a maximum of 500 ns. It is apparentthat while, in the case of GeTe, we have a good crystallizationfor all pulse lengths (with a minimum resistance contrast oftwo orders of magnitude), in the case of GST, the final resistivevalue is much more sensitive to pulse length and amplitude.
In other words, assuming a RESET/SET contrast criterion oftwo decades, the SET state in the GeTe device is achieved in30 ns, while the SET state in the GST device is obtained in500 ns: GeTe allows a gain of more than one decade in ener-gy per bit with respect to GST. In agreement with amorphous/
Fig. 2. Programming performance for (a) GST and (b) GeTe devices withdifferent SET times; each SET pulse is preceded by a fixed RESET pulse(IRESET = 30 mA/tRESET = 30 ns). Programming currents are high (dueto large W plug). Values should be considered as terms of comparison betweenGST and GeTe. Note that GeTe shows much faster SET operation with a highercontrast between RESET and SET states than that of GST. (Inset) Schematicsof the fabricated device.
Fig. 3. Endurance characteristics for (a)–(c) GST and (b)–(d) GeTe. In(a)–(b), a long SET pulse is used to program the cells. In (c)–(d), a short SETpulse of 30 ns is used. Note that no intelligent algorithm (i.e., variable numberof pulses to maintain a fixed RESET/SET contrast) is used to cycle the cellsand that IRESET = 26 mA and ISET = 18 mA for both GST and GeTe.
crystalline contrast on full-sheet depositions (Section II-A),note that the maximum achievable contrast is approximatelythree orders of magnitude for GeTe and approximately two forGST, with the crystalline state being much more conductive forGeTe. Moreover, the fast resistivity drop noticed in Section II-Afor GeTe could justify the very fast transition between RESETand SET states [Fig. 2(b)].
2) Endurance Characteristics: The results of the endurancetest, with different SET times, are shown in Fig. 3. In agreement
490 IEEE ELECTRON DEVICE LETTERS, VOL. 31, NO. 5, MAY 2010
Fig. 4. Data retention for the nine devices at (a) 160 ◦C for GeTe and(b) 125 ◦C for GST after identical RESET pulses IRESET = 30 mA/tRESET = 60 ns. In the inset, the geometric average of the resistance evolutionof GeTe and GST is the same during crystallization. Note that the reportedretention experiments are performed for the same ratio T/Tm, where Tm isthe material melting temperature equal to 903 and 996 K for GST and GeTe,respectively.
with the results shown in Fig. 2, for a 200-ns pulse, GSTmaintains a resistive contrast of about two decades betweenRESET and SET states [Fig. 3(a)] while GeTe shows morethan three decades [Fig. 3(b)]. In both cases, no cell failureis apparent up to 1 × 106 cycles. In the case of a short SETpulse, it is apparent that GST displays a narrower resistancewindow which closes as the cycle count increases [Fig. 3(c)].On the contrary, GeTe shows a very good endurance up to1 × 105 cycles with SET/RESET pulses as short as 30 ns[Fig. 3(d)].
3) Data Retention Characteristics: In Fig. 4, we have rep-resented the data-retention characteristics for GeTe and GST. Inthese experiments, the chuck is heated up to high temperature,and then, the sequence (RESET pulse, plus repeated resistancemeasurements) is performed on nine cells. As shown in theinset, similar average retention behaviors (i.e., resistance lossin the same timescale) are obtained at 160 ◦C for GeTe and at125 ◦C for GST. These results suggest that GeTe has a bet-ter thermal stability than GST, in agreement with the highercrystallization temperature measured on full-sheet deposition(Tc ∼ 185 ◦C and 145 ◦C, respectively [5], [6]). A moreuniform behavior in GST than that in GeTe devices seems toappear and should be better investigated. At the onset of thecrystallization process, the crystallization speed is much higherin GeTe than in GST. The same effect has been noticed inSection II-A.
This behavior can be related to the different interplays be-tween nucleation and growth in the crystallization process,which is less known for GeTe [6], [8]–[10]. Moreover, thestronger contrast, between crystalline and amorphous resistiv-ity, in GeTe than that in GST (shown in Section II-A) couldjustify a faster drop in cell resistance as soon as a crystallinepath is created through the amorphous spot.
III. CONCLUSION
In this letter, we have presented the performance and relia-bility characteristics of GeTe-based PCM cells and comparedthem to data obtained on GST PCM cells. Results show that thefollowing are true.
1) GeTe devices show very fast program characteristics (inagreement with the literature [7]), allowing a gain ofmore than one decade in energy per bit with respectto GST, for SET operations. The RESET/SET contrastis approximately three decades for GeTe, while it isapproximately two decades for GST, with the SET statebeing more conductive in GeTe than in GST.
2) GeTe devices allow stable endurance up to 1 × 105 cycleswith RESET/SET pulses as short as 30 ns.
3) Similar data-retention characteristics are shown for GeTeat 160 ◦C and for GST at 125 ◦C. This result agrees withthe difference in crystallization temperatures measuredon full-sheet deposition and noticed in [5] and [6].
As a conclusion, this letter sheds new light on GeTe as analternative material to GST in PCMs, eventually allowing us toaddress applications where programming speed and bandwidthare requested (i.e., caching) or where strict requirements on dataretention at high temperatures exist (i.e., embedded NVMs).
REFERENCES
[1] S. Lai, “Non-volatile memory technologies: The quest for ever lowercost,” in IEDM Tech. Dig., Dec. 2008, pp. 1–6.
[2] B. Gleixner, F. Pellizzer, and R. Bez, “Reliability characterization of phasechange memory,” in Proc. Eur. Phase-Change Ovonics Symp., 2009,pp. 135–139.
[3] T. Morikawa, K. Kurotsuchi, M. Kinoshita, N. Matsuzaki, Y. Matsui,Y. Fuiisaki, S. Hanzawa, A. Kotabe, M. Terao, H. Moriya, T. Iwasaki,M. Matsuoka, F. Nitta, M. Moniwa, T. Koga, and N. Takaura, “DopedIn–Ge–Te phase change memory featuring stable operation and good dataretention,” in IEDM Tech. Dig., Dec. 2007, pp. 307–310.
[4] 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, andM. Moniwa, “Oxygen-doped GeSbTe phase-change memory cells featur-ing 1.5 V/100 μA standard 0.13 μm CMOS operations,” in IEDM Tech.Dig., Dec. 2005, pp. 738–741.
[5] A. Fantini, L. Perniola, M. Armand, J. F. Nodin, V. Sousa, A. Persico,J. Cluzel, C. Jahan, S. Maitrejean, S. Lhostis, A. Roule, C. Dressler,G. Reimbold, B. DeSalvo, P. Mazoyer, D. Bensahel, and F. Boulanger,“Comparative assessment of GST and GeTe materials for application toembedded phase-change memory devices,” in Proc. Int. Memory Work-shop, 2009, pp. 66–67.
[6] S. Raoux, H.-Y. Cheng, M. A. Caldwell, and H.-S. P. Wong, “Crystalliza-tion times of Ge–Te phase change materials as a function of composition,”Appl. Phys. Lett., vol. 95, no. 7, p. 071 910, Jul. 2009.
[7] G. Bruns, P. Merkelbach, C. Schlockermann, M. Salinga, M. Wuttig,T. D. Happ, J.B. Philipp, and M. Kund, “Nanosecond switching in GeTephase change memory cells,” Appl. Phys. Lett., vol. 95, no. 4, p. 043 108,Jul. 2009.
[8] J. H. Coombs, A. P. J. M. Jongenelis, W. Van Es-Spiekman, andB. A. J. Jacobs, “Laser-induced crystallization phenomena in GeTe-basedalloys. I. Characterization of nucleation and growth,” J. Appl. Phys.,vol. 78, no. 8, pp. 4906–4917, Oct. 1995.
[9] J. H. Coombs, A. P. J. M. Jongenelis, W. Van Es-Spiekman, andB. A. J. Jacobs, “Laser-induced crystallization phenomena in GeTe-basedalloys. II. Composition dependence of nucleation and growth,” J. Appl.Phys., vol. 78, no. 8, pp. 4918–4928, Oct. 1995.
[10] J. H. Coombs, A. P. J. M. Jongenelis, W. Van Es-Spiekman, andB. A. J. Jacobs, “Laser-induced crystallization phenomena in GeTe-basedalloys. III. GeTeSe alloys for a CD compatible erasable disk,” J. Appl.Phys., vol. 79, no. 11, pp. 8349–8356, Jun. 1996.
