- Email: [email protected]
Conduction and stability of holmium titanium oxide thin films grown by atomic layer deposition
Thin Solid Films 591 (2015) 55–59
Contents lists available at ScienceDirect
Thin Solid Films journal homepage: www.elsevier.com/locate/tsf
Conduction and stability of holmium titanium oxide thin films grown by atomic layer deposition H. Castán a,⁎, H. García a, S. Dueñas a, L. Bailón a, E. Miranda b, K. Kukli c,d, M. Kemell c, M. Ritala c, M. Leskelä c a
Department of Electronic, University of Valladolid, 47011 Valladolid, Spain Departament d'Enginyería Electrònica, Universitat Autónoma de Barcelona, 08193 Bellaterra, Spain Department of Chemistry, University of Helsinki, FI-00014 Helsinki, Finland d Institute of Physics, University of Tartu, EE-50411,Tartu, Estonia b c
a r t i c l e
i n f o
Article history: Received 27 January 2015 Received in revised form 3 August 2015 Accepted 14 August 2015 Available online 18 August 2015 Keywords: Holmium titanium oxide Resistive switching Oxide breakdown
a b s t r a c t Holmium titanium oxide (HoTiOx) thin films of variable chemical composition grown by atomic layer deposition are studied in order to assess their suitability as dielectric materials in metal–insulator–metal electronic devices. The correlation between thermal and electrical stabilities as well as the potential usefulness of HoTiOx as a resistive switching oxide are also explored. It is shown that the layer thickness and the relative holmium content play important roles in the switching behavior of the devices. Cycled current–voltage measurements showed that the resistive switching is bipolar with a resistance window of up to five orders of magnitude. In addition, it is demonstrated that the post-breakdown current–voltage characteristics in HoTiOx are well described by a power-law model in a wide voltage and current range which extends from the soft to the hard breakdown regimes. © 2015 Elsevier B.V. All rights reserved.
1. Introduction Ho2O3, like other oxides of the lanthanide series such as Gd2O3, has been proposed as a dielectric material for advanced electron devices. This material exhibits a moderate permittivity value (k ≈ 12) as well as a large band gap (5.3 eV) [1,2]. Ho-doped TiO2 and HoTiO3 are also high permittivity (high-k) materials and possible candidates to replace SiO2 in field-effect transistors [3]. Although TiO2 is considered a good candidate as a dielectric layer in metal–insulator–metal (MIM) capacitors for dynamic random access memories (DRAM) [4] because of its high permittivity value, its major drawback is its small band gap/low barrier height and hence its large leakage current when biased. However, when combined with a larger band gap material, the leakage current reduces significantly: Al-doped TiO2 MIM capacitors with an equivalent oxide thickness of 0.56 nm with leakage current of around 1 × 10−8 A cm−2 have been fabricated [5]. This means a current density decrease of 106 times due to Al doping. As it is known, in order to overcome the DRAM limitations, such as the high power consumption and stored charge volatility, new kinds of memories are currently under investigation. In particular, resistive random access memories (RRAMs), which are based on a two-terminal device structure that stores data in the form of a reversible resistance change, seem to be very promising [6,7]. Depending on the particular features of the ⁎ Corresponding author. E-mail address: [email protected] (H. Castán).
http://dx.doi.org/10.1016/j.tsf.2015.08.027 0040-6090/© 2015 Elsevier B.V. All rights reserved.
current–voltage (I–V) characteristic, the switching behavior can be unipolar (the switching direction depends on the magnitude of the applied voltage but not on its polarity) or bipolar (the change of the state can be obtained by applying voltages with specific polarity). Resistive switching (RS) phenomena have been found in various types of binary oxides including TiO2, ZrO2, HfOx, etc. [8–12]. In fact, TiO2 is one of the most studied materials for RS applications due to its simple fabrication at low process temperature and mass production compatibility. As it has been said before, the insertion of a thin Al2O3 barrier layer helps to overcome its high operational currents. B. Hudec, et al. [13] systematically studied the effect of this layer thickness on the RS behavior of stacks with 20 nm of TiO2 covered with an Al2O3 layer. They found that an Al2O3 layer of a certain thickness, around 3–4 nm, is essential to stabilize the RS parameters. Moreover, no RS could be observed in the single layer TiO2 due to its leaky character. Other authors have reported RS behavior in structures such as Pt/TiO2/Pt [8,14,15], but large fluctuations in the I–V sweep curves exist during the reset operations, which are one of the big hurdles that the RS memory has to overcome to become a viable device [12]. Recently, a model has been proposed to explain these large fluctuations [16], that might have their origin in the stochastic nature of the set processes during the repeated RS via the I–V sweeps. In this work, holmium titanium oxide thin films of different chemical compositions grown by atomic layer deposition have been electrically characterized. A variety of MIM structures have been fabricated in order to assess the suitability of these dielectric films for microelectronics
56
H. Castán et al. / Thin Solid Films 591 (2015) 55–59
use. Special attention has been paid to the post-breakdown I–V characteristics because of its connection with the resistive switching effect. 2. Experimental details The films containing Ho were grown in a commercial flow-type hotwall reactor F120 (ASM Microchemistry, Ltd.) [17]. Table 1 summarizes the samples studied in this work. Ho2O3 was grown using Ho(thd)3 (thd = 2,2,6,6-tetramethyl-3,5-heptanedionato) and ozone as holmium and oxygen precursors respectively. TiO2 films were grown using titanium-isopropoxide, Ti(OCH(CH3)2)4 and ozone as precursors. Deposition temperature was 300 °C for both constituent oxides. TiO2:Ho2O3 films were grown as stacks of alternating layers of TiO2 and Ho2O3. The thicknesses of the films were evaluated by X-ray reflectometry. The ratios of the different metals were calculated by an energy dispersive X-ray spectrometer. The growth rate of Ho2O3 is lower than that of TiO2, and it was thus necessary to apply, at least, two Ho2O3 cycles for each TiO2 cycle in order to achieve HoTiO3 or Ho2Ti2O7 stoichiometry. Selected samples were annealed under nitrogen atmosphere for 30 min at 600 or 800 °C. More detailed description can be found in an earlier study devoted to the structure and morphology of these samples [18]. Fig. 1 shows grazing incidence diffraction patterns, taken from K. Kukli, et al. [18], corresponding to as-deposited Ho2O3 film (Fig. 1(a)), and annealed HoTiOx films, with Ho:Ti = 0.54 and 0.48 (Fig. 1(b) and (c), respectively). The Ho2O3 films became crystallized in the asdeposited state, whereas in the HoTiOx films the crystallization required annealing. In order to make MIM structures, the TiO2:Ho2O3 films were deposited on low resistivity silicon substrates covered with 10 nm thick TiN layers. Al/Ti top electrodes (N 100/30 nm thick) were e-beam evaporated through a shadow mask. After the top electrode deposition, samples were not additionally annealed. The electrical measurements were carried out with two different capacitor sizes (2.04 × 10−3 and 0.52 × 10−3 cm2). A permittivity value of around 25 was obtained for the annealed 25.1 nm-thick HoTiOx, with Ho:Ti atomic ratio of 1.5 [19]. The permittivity increased after annealing, when the samples became crystallized [18]. The highest permittivity values (around 40), correspond to the lowest Ho:Ti values (0.54 and 0.48). I–V measurements were carried out using an HP4155B semiconductor parameter analyzer, whereas a Keithley 4200-SCS semiconductor characterization system was used to obtain capacitance–voltage (C–V) data. I–V measurements revealed the existence of reversible bipolar resistive switching in some of the holmium oxide based devices. As it will be shown in the next section, a clear influence of holmium content and annealing on the C–V linearity has been found. 3. Results and discussion Fig. 2 shows room temperature C–V curves corresponding to HoTiOx-based MIM capacitors with thicknesses of 7.7, 18.7 and 25.1 nm and Ho:Ti atomic ratios of 1.4, 0.9 and 1.5, respectively. As shown in this figure, the C–V curves are asymmetrical. Capacitance increases for positive voltages and decreases for negative ones. This effect which is relatively weak in the as-deposited samples becomes much more pronounced in the annealed ones. C–V curves are linear for the as-deposited samples and the capacitance values scale according to the dielectric film thickness, i.e. a thinner insulator is associated with a Table 1 HoTiOx samples fabricated for this study. Material
Metal compounds atomic ratio Ho:Ti
d (nm)
Annealing (N2, 30′)
HoTiOx HoTiOx HoTiOx HoTiOx HoTiOx Ho2O3
0.48 0.54 0.9 1.4 1.5 –
23.8 27.5 18.7 7.7 25.1 22.7
600 °C 600 °C/800 °C 600 °C/800 °C 800 °C 800 °C 600 °C
Fig. 1. Grazing incidence diffraction patterns of as-deposited Ho2O3 films (a), and annealed HoTiOx films, with Ho:Ti = 0.54 (b), and 0.48 (c). (Taken from K. Kukli, et al. [18]).
higher capacitance value. However, notice that the C–V curves of the annealed samples are not linear at all. Capacitance is at a minimum value for negative voltages (lower than for the as-deposited samples), then strongly increases up to a certain maximum and then again decreases for positive voltages to an even lower value. Two points are worth mentioning: first, the thickest sample reaches capacitance values similar to those at positive voltage values and, second, the thinnest sample does not reach so high values but returns to values similar to those exhibited by the as-deposited sample. Moreover, maximum capacitance occurs at positive voltages (0.3 V) for the 18.7 and 25.1 nm thick samples, and at a negative voltage (− 0.4 V) for the 7.7 nm thick sample. Current conduction measurements show parallel trends: the I–V curves for the same samples are plotted in Fig. 3. These curves were measured on fresh devices, i.e. in samples that were not previously biased. The
H. Castán et al. / Thin Solid Films 591 (2015) 55–59
57
Fig. 2. Capacitance–voltage curves measured at room temperature corresponding to HoTiOx-based MIM structures with Ho:Ti atomic ratios of 0.9, 1.4, and 1.5.
measurements were carefully carried out to avoid resistance switching: the voltage was first decreased from 0 to −1 V and then the current was recorded during its slow increment (sweep rate = 5 mV/s) up to +1 V. In all the cases, the annealed samples show higher current values than the as-deposited ones indicating some degradation of the insulating layer. In addition, the thinnest sample shows the poorest current behavior even before annealing. Holmium content also plays an important role in the linearity of capacitance of the as-deposited samples. Fig. 4(a) and (b) shows C–V curves for the as-deposited and annealed samples with the lowest values of holmium content (Ho:Ti = 0.54 and 0.48). These curves are all clearly non-linear, even for the as-deposited samples. In addition, in all cases the capacitance shows a maximum at negative values (about − 0.5 V) and reaches lower values at positive bias voltages. In Fig. 4(b) we see that the annealing process reduces the capacitance value, just in the opposite direction observed in the previous samples. As for the current values, one can see in Fig. 5 that these samples show poor fresh I–V curves and, once again, the sample with higher capacitance (as-deposited 23.8 nm-thick, Ho:Ti = 0.48) is the most leaky. These current values are too high for device application. On the other hand, it has been widely reported that defective dielectrics exhibit resistance switching effect that can be exploited to fabricate nonvolatile memory devices [6,7]. We have detected this behavior in the three samples with the lowest current density values (Fig. 3): for 18.7 (both as-deposited and annealed) and 25.1 nm-thick (as deposited) HoTiOx-based MIM samples, with Ho:Ti atomic ratio of 0.9 and 1.5, respectively, and we have also observed resistive switching in the
Fig. 3. Current density of pristine samples: HoTiOx-based MIM structures with Ho:Ti atomic ratios of 0.9, 1.4, and 1.5.
Fig. 4. Capacitance–voltage curves measured at room temperature corresponding to annealed HoTiOx-based MIM structures with Ho:Ti atomic ratios of 0.54 and 0.48 (a), and for the annealed and as-deposited samples with Ho:Ti atomic ratio of 0.48 (b).
sample with no titanium content (22.7 nm-thick stoichiometric Ho2O3). Resistive switching was not observed in samples with the lowest Ho:Ti ratio values, i.e. for 27.5 and 23.8 nm-thick HoTiOx based-MIM samples with Ho:Ti atomic ratio of 0.54 and 0.48, respectively. The thinnest sample (7.7 nm-thick HoTiOx sample with Ho:Ti atomic ratio of 1.4) does not show resistive switching behavior either. Therefore, both
Fig. 5. Current density of pristine samples: HoTiOx-based MIM structures with Ho:Ti atomic ratios of 0.54 and 0.48.
58
H. Castán et al. / Thin Solid Films 591 (2015) 55–59
Fig. 7. (a) Bipolar resistive switching cycles corresponding to the as-deposited 18.7 nm-thick HoTiOx with a Ho:Ti atomic ratio of 0.9 based MIM sample. (b) Endurance plot of current at 0.5 V positive and negative voltages.
Fig. 6. Three resistive-switching cycle plots for MIM samples of as-deposited 25.1 nm-thick HoTiOx with a Ho:Ti atomic ratio of 1.5 (a), and a pure 22.7 nm thick H2O3 annealed sample (b) films.
thickness and holmium content seem to be the key parameters to obtain useful devices. In Fig. 6(a) and (b), three resistive switching cycles for HoTiOx (25.1 nm-thick, Ho:Ti = 1.5) and Ho2O3 (22.7 nm-thick) samples, respectively, are plotted. In this figure one can see that the I–V cycles show bipolar resistive switching, since the bias voltage polarity has to be changed for achieving transitions between states. In the case of the 25.1 nm-thick layer, there are up to five orders of magnitude between the low resistance state (LRS) and the high resistance state (HRS) values, both measured at + 1 V. As the leakage current increases, the change in resistance decreases; in fact, the low resistivity current value is very similar for all the devices. In the Ho2O3 MIM sample, the LRS current measured at +1 V decreased whereas resistance reaches a value of around 60 MΩ, a really adequate value to achieve low power consumption. We have further investigated the RS behavior of our devices by performing 35 consecutive I–V cycles on the as-deposited 18.7 nm-thick HoTiOx-based MIM sample with Ho:Ti = 0.9 (see Fig. 7(a)). We observe larger current variability for HRS than for LRS, both at positive and negative voltages. This behavior can be more clearly observed in Fig. 7(b) in which the current values are plotted between +0.5 and −0.5 V for both states. This can be understood taken into account the size of the filamentary path and its susceptibility to local changes in the defects configurations. Very convenient resistance values are obtained for LRS (around 1 MΩ) and HRS (lower than 100 MΩ). However, the endurance is still poor: after the cycle number 30 the current at the
low resistance state gradually decays down to the HRS. Therefore, although these samples are promising and do show RS effect, more work has to be made in order to make them good challengers for RRAM device fabrication. To conclude, a remarkable feature exhibited by the post-breakdown HoTiOx I–V characteristic is illustrated in Fig. 8. This figure shows that the curves can be nicely fitted using a power-law model I = aVb (Fig. 8) in a very wide voltage and current range. This behavior is typical of the soft breakdown failure mode (HRS) in ultrathin (b5 nm) SiO2 layers [20] but has also been reported for a variety of high-k dielectrics [21–23]. In the present case, the soft breakdown mode occurs in a
Fig. 8. Experimental (colored lines) and modeled (black lines) I–V curves corresponding to the annealed 18.7 nm-thick HoTiOx with a Ho:Ti atomic ratio of 0.9 based MIM sample. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
H. Castán et al. / Thin Solid Films 591 (2015) 55–59
relatively thick (around 18.7 nm) film. Earlier reports indicate that RS has been observed in even thicker TiO2 films [24]. Interestingly, in HoTiOx the power-law model can be applied up to the hard breakdown mode (LRS) with a small deviation at the largest current values in the positive bias range because of the potential drop in the semiconductor substrate (inversion condition). 4. Conclusions Atomic layer deposited films of holmium titanium oxide (HoTiOx) have been studied in order to explore their suitability to be exploited for conventional gate dielectric material and nonvolatile memory applications. HoTiOx MIM structures clearly exhibit resistive switching effect. However, the number of LRS to HRS switching cycles is still low and future work must be done in order to improve the endurance of these promising structures. It was also shown that the post-breakdown I–V characteristics can be fitted using a power-law model both in the LRS and HRS regimes. This extends the validity of this model to the hard breakdown regime. Acknowledgments The study has been supported by the Spanish TEC2014 under grant No. 52152-C3-3-R, and by the Finnish Centre of Excellence in Atomic Layer Deposition. E. Miranda acknowledges the Ministerio de Economía y Competitividad, Spain (TEC 2012-32305) with support of federal funds. References [1] T.M. Pan, M.D. Huang, Structural properties and sensing characteristics of high-k Ho2O3 sensing film-based electrolyte–insulator–semiconductor, Mater. Chem. Phys. 129 (2011) 919–924. [2] G. Darbandy, F. Lime, A. Cerdeira, M. Estrada, I. Garduño, B. Iñiguez, Study of potential high-k dielectric for UTB SOI MOSFETS using analytical modeling of the gate tunneling leakage, Semicond. Sci. Technol. 26 (2011) 115002. [3] T.M. Pan, M.D. Huang, C.W. Lin, M.H. Wu, Development of high-k HoTiO3 sensing membrane for pH detection and glucose biosensing, Sensors Actuators B Chem. 144 (2010) 139–145. [4] M. Seo, S.H. Rha, S.K. Kim, J.H. Han, W. Lee, S. Han, C.S. Hwang, The mechanism for the suppression of leakage current in high dielectric TiO2 thin films by adopting ultra-thin HfO2 films for memory application, J. Appl. Phys. 110 (2011) 024105. [5] S.K. Kim, S.W. Lee, J.H. Han, B. Lee, S. Han, C.S. Hwang, Capacitors with an equivalent oxide thickness of b0.5 nm for nanoscale electronic semiconductor memory, Adv. Funct. Mater. 20 (2010) 2989–3003. [6] R. Waser, M. Aono, Nanoionics-based resistive switching memories, Nat. Mater. 6 (2007) 833–840.
59
[7] D.S. Jeong, R. Thomas, R.S. Katiyar, J.F. Scott, H. Kohlstedt, A. Petraru, C.S. Hwang, Emerging memories: resistive switching mechanisms and current status, Rep. Prog. Phys. 75 (2012) 076502. [8] C. Rohde, B.J. Choi, D.S. Jeong, S. Choi, J.S. Zhao, C.S. Hwang, Identification of a determining parameter for resistive switching of TiO2 thin films, Appl. Phys. Lett. 86 (2005) 262907. [9] D. Lee, H. Choi, H. Sim, D. Choi, H. Hwang, M.J. Lee, S.A. Seo, I.K. Yoo, Resistance switching of the nonstoichiometric zirconium oxide for nonvolatile memory applications, IEEE Electron Device Lett. 26 (2005) 719–721. [10] S. Spiga, A. Lamperti, C. Wiemer, M. Perego, E. Cianci, G. Tallarida, H.L. Lu, M. Alia, F.G. Volpe, M. Fanciulli, Resistance switching in amorphous and crystalline binary oxides grown by electron beam evaporation and atomic layer deposition, Microelectron. Eng. 85 (2008) 2414–2419. [11] Y. Li, S. Long, M. Zhang, Q. Liu, L. Shao, S. Zhang, Y. Wang, Q. Zuo, S. Liu, M. Liu, Resistive switching properties of Au/ZrO2/Ag structure for low-voltage nonvolatile memory applications, IEEE Electron Device Lett. 31 (2010) 117–119. [12] D. Acharyya, A. Hazra, P. Bhattacharyya, A journey towards reliability improvement of TiO2 based resistive random acces memory: a review, Microelectron. Reliab. 54 (2014) 541–560. [13] B. Hudec, A. Paskaleva, P. Jancovic, J. Dérer, J. Fedor, A. Rosová, E. Dobrocka, K. Fröhlich, Resistive switching in TiO2-based metal–insulator–metal structures with Al2O3 barrier layer at the metal/dielectric interface, Thin Solid Films 563 (2014) 10–14. [14] K.M. Kim, B.J. Choi, C.S. Hwang, Localized switching mechanism in resistive switching of atomic layer deposited TiO2 thin films, Appl. Phys. Lett. 90 (2007) 242906. [15] K.M. Kim, B.J. Choi, Y.C. Shin, S. Choi, C.S. Hwang, Anode-interface localized filamentary mechanism in resistive switching of TiO2 thin films, Appl. Phys. Lett. 91 (2007) 012907. [16] K.M. Kim, T.H. Park, C.S. Hwang, Dual conical conducting filament model in resistance switching TiO2 thin films, Sci. Rep. 5 (2015) 7844, http://dx.doi.org/10.1038/ srep07844. [17] T. Suntola, Atomic layer epitaxy, Thin Solid Films 216 (1992) 84–89. [18] K. Kukli, M. Kemell, M.C. Dimri, E. Puukilainen, A. Tamm, R. Stern, M. Ritala, M. Leskelä, Holmium titanium oxide thin films grown by atomic layer deposition, Thin Solid Films 565 (2014) 261–266. [19] H. García, H. Castán, S. Dueñas, E. Pérez, L.A. Bailón, K. Kukli, M. Ritala, M. Leskelä, Resistive switching behavior and electrical properties of TiO2:Ho2O3 and HoTiOx based MIM capacitors, Mater. Res. Soc. Symp. Proc. 1691 (2014), http://dx.doi.org/ 10.1557/opl.2014.598. [20] E. Miranda, J. Suñé, R. Rodríguez, M. Nafría, X. Aymerich, A function-fit model for the soft breakdown failure mode, IEEE Electron Device Lett. 20 (1999) 265–267. [21] M. Houssa, P.W. Mertens, M.M. Heyns, J.S. Jeon, A. Halliyal, B. Ogle, Soft breakdown in very thin Ta2O5 gate dielectric layers, Solid State Electron. 44 (2000) 521–525. [22] E. Miranda, J. Martin-Martinez, E. O'Connor, G. Hughes, P. Casey, K. Cherkaoui, S. Monaghan, R. Long, D. O'Connell, P.K. Hurley, Effects of the electrical stress on the conduction characteristics of metal gate/MgO/InP stacks, Microelectron. Reliab. 49 (2009) 1052–1055. [23] M.B. González, J.M. Rafí, O. Beldarrain, M. Zabala, F. Campabadal, Charge trapping and electrical degradation in atomic layer deposited Al2O3 films, Microelectron. Eng. 109 (2013) 57–59. [24] B. Hudec, M. Hranai, K. Hušekova, J. Aarik, A. Tarre, K. Fröhlich, RuO2/TiO2 based MIM capacitors for DRAM application, The 8th Int. Conf. on Advanc. Semicond. Dev. and Microsystems (ASDAM 2010), Smolenice, Slovakia, 25–27 Oct. 2010, IEEE, Piscataway, NJ, USA 2010, pp. 255–258.
Contents lists available at ScienceDirect
Thin Solid Films journal homepage: www.elsevier.com/locate/tsf
Conduction and stability of holmium titanium oxide thin films grown by atomic layer deposition H. Castán a,⁎, H. García a, S. Dueñas a, L. Bailón a, E. Miranda b, K. Kukli c,d, M. Kemell c, M. Ritala c, M. Leskelä c a
Department of Electronic, University of Valladolid, 47011 Valladolid, Spain Departament d'Enginyería Electrònica, Universitat Autónoma de Barcelona, 08193 Bellaterra, Spain Department of Chemistry, University of Helsinki, FI-00014 Helsinki, Finland d Institute of Physics, University of Tartu, EE-50411,Tartu, Estonia b c
a r t i c l e
i n f o
Article history: Received 27 January 2015 Received in revised form 3 August 2015 Accepted 14 August 2015 Available online 18 August 2015 Keywords: Holmium titanium oxide Resistive switching Oxide breakdown
a b s t r a c t Holmium titanium oxide (HoTiOx) thin films of variable chemical composition grown by atomic layer deposition are studied in order to assess their suitability as dielectric materials in metal–insulator–metal electronic devices. The correlation between thermal and electrical stabilities as well as the potential usefulness of HoTiOx as a resistive switching oxide are also explored. It is shown that the layer thickness and the relative holmium content play important roles in the switching behavior of the devices. Cycled current–voltage measurements showed that the resistive switching is bipolar with a resistance window of up to five orders of magnitude. In addition, it is demonstrated that the post-breakdown current–voltage characteristics in HoTiOx are well described by a power-law model in a wide voltage and current range which extends from the soft to the hard breakdown regimes. © 2015 Elsevier B.V. All rights reserved.
1. Introduction Ho2O3, like other oxides of the lanthanide series such as Gd2O3, has been proposed as a dielectric material for advanced electron devices. This material exhibits a moderate permittivity value (k ≈ 12) as well as a large band gap (5.3 eV) [1,2]. Ho-doped TiO2 and HoTiO3 are also high permittivity (high-k) materials and possible candidates to replace SiO2 in field-effect transistors [3]. Although TiO2 is considered a good candidate as a dielectric layer in metal–insulator–metal (MIM) capacitors for dynamic random access memories (DRAM) [4] because of its high permittivity value, its major drawback is its small band gap/low barrier height and hence its large leakage current when biased. However, when combined with a larger band gap material, the leakage current reduces significantly: Al-doped TiO2 MIM capacitors with an equivalent oxide thickness of 0.56 nm with leakage current of around 1 × 10−8 A cm−2 have been fabricated [5]. This means a current density decrease of 106 times due to Al doping. As it is known, in order to overcome the DRAM limitations, such as the high power consumption and stored charge volatility, new kinds of memories are currently under investigation. In particular, resistive random access memories (RRAMs), which are based on a two-terminal device structure that stores data in the form of a reversible resistance change, seem to be very promising [6,7]. Depending on the particular features of the ⁎ Corresponding author. E-mail address: [email protected] (H. Castán).
http://dx.doi.org/10.1016/j.tsf.2015.08.027 0040-6090/© 2015 Elsevier B.V. All rights reserved.
current–voltage (I–V) characteristic, the switching behavior can be unipolar (the switching direction depends on the magnitude of the applied voltage but not on its polarity) or bipolar (the change of the state can be obtained by applying voltages with specific polarity). Resistive switching (RS) phenomena have been found in various types of binary oxides including TiO2, ZrO2, HfOx, etc. [8–12]. In fact, TiO2 is one of the most studied materials for RS applications due to its simple fabrication at low process temperature and mass production compatibility. As it has been said before, the insertion of a thin Al2O3 barrier layer helps to overcome its high operational currents. B. Hudec, et al. [13] systematically studied the effect of this layer thickness on the RS behavior of stacks with 20 nm of TiO2 covered with an Al2O3 layer. They found that an Al2O3 layer of a certain thickness, around 3–4 nm, is essential to stabilize the RS parameters. Moreover, no RS could be observed in the single layer TiO2 due to its leaky character. Other authors have reported RS behavior in structures such as Pt/TiO2/Pt [8,14,15], but large fluctuations in the I–V sweep curves exist during the reset operations, which are one of the big hurdles that the RS memory has to overcome to become a viable device [12]. Recently, a model has been proposed to explain these large fluctuations [16], that might have their origin in the stochastic nature of the set processes during the repeated RS via the I–V sweeps. In this work, holmium titanium oxide thin films of different chemical compositions grown by atomic layer deposition have been electrically characterized. A variety of MIM structures have been fabricated in order to assess the suitability of these dielectric films for microelectronics
56
H. Castán et al. / Thin Solid Films 591 (2015) 55–59
use. Special attention has been paid to the post-breakdown I–V characteristics because of its connection with the resistive switching effect. 2. Experimental details The films containing Ho were grown in a commercial flow-type hotwall reactor F120 (ASM Microchemistry, Ltd.) [17]. Table 1 summarizes the samples studied in this work. Ho2O3 was grown using Ho(thd)3 (thd = 2,2,6,6-tetramethyl-3,5-heptanedionato) and ozone as holmium and oxygen precursors respectively. TiO2 films were grown using titanium-isopropoxide, Ti(OCH(CH3)2)4 and ozone as precursors. Deposition temperature was 300 °C for both constituent oxides. TiO2:Ho2O3 films were grown as stacks of alternating layers of TiO2 and Ho2O3. The thicknesses of the films were evaluated by X-ray reflectometry. The ratios of the different metals were calculated by an energy dispersive X-ray spectrometer. The growth rate of Ho2O3 is lower than that of TiO2, and it was thus necessary to apply, at least, two Ho2O3 cycles for each TiO2 cycle in order to achieve HoTiO3 or Ho2Ti2O7 stoichiometry. Selected samples were annealed under nitrogen atmosphere for 30 min at 600 or 800 °C. More detailed description can be found in an earlier study devoted to the structure and morphology of these samples [18]. Fig. 1 shows grazing incidence diffraction patterns, taken from K. Kukli, et al. [18], corresponding to as-deposited Ho2O3 film (Fig. 1(a)), and annealed HoTiOx films, with Ho:Ti = 0.54 and 0.48 (Fig. 1(b) and (c), respectively). The Ho2O3 films became crystallized in the asdeposited state, whereas in the HoTiOx films the crystallization required annealing. In order to make MIM structures, the TiO2:Ho2O3 films were deposited on low resistivity silicon substrates covered with 10 nm thick TiN layers. Al/Ti top electrodes (N 100/30 nm thick) were e-beam evaporated through a shadow mask. After the top electrode deposition, samples were not additionally annealed. The electrical measurements were carried out with two different capacitor sizes (2.04 × 10−3 and 0.52 × 10−3 cm2). A permittivity value of around 25 was obtained for the annealed 25.1 nm-thick HoTiOx, with Ho:Ti atomic ratio of 1.5 [19]. The permittivity increased after annealing, when the samples became crystallized [18]. The highest permittivity values (around 40), correspond to the lowest Ho:Ti values (0.54 and 0.48). I–V measurements were carried out using an HP4155B semiconductor parameter analyzer, whereas a Keithley 4200-SCS semiconductor characterization system was used to obtain capacitance–voltage (C–V) data. I–V measurements revealed the existence of reversible bipolar resistive switching in some of the holmium oxide based devices. As it will be shown in the next section, a clear influence of holmium content and annealing on the C–V linearity has been found. 3. Results and discussion Fig. 2 shows room temperature C–V curves corresponding to HoTiOx-based MIM capacitors with thicknesses of 7.7, 18.7 and 25.1 nm and Ho:Ti atomic ratios of 1.4, 0.9 and 1.5, respectively. As shown in this figure, the C–V curves are asymmetrical. Capacitance increases for positive voltages and decreases for negative ones. This effect which is relatively weak in the as-deposited samples becomes much more pronounced in the annealed ones. C–V curves are linear for the as-deposited samples and the capacitance values scale according to the dielectric film thickness, i.e. a thinner insulator is associated with a Table 1 HoTiOx samples fabricated for this study. Material
Metal compounds atomic ratio Ho:Ti
d (nm)
Annealing (N2, 30′)
HoTiOx HoTiOx HoTiOx HoTiOx HoTiOx Ho2O3
0.48 0.54 0.9 1.4 1.5 –
23.8 27.5 18.7 7.7 25.1 22.7
600 °C 600 °C/800 °C 600 °C/800 °C 800 °C 800 °C 600 °C
Fig. 1. Grazing incidence diffraction patterns of as-deposited Ho2O3 films (a), and annealed HoTiOx films, with Ho:Ti = 0.54 (b), and 0.48 (c). (Taken from K. Kukli, et al. [18]).
higher capacitance value. However, notice that the C–V curves of the annealed samples are not linear at all. Capacitance is at a minimum value for negative voltages (lower than for the as-deposited samples), then strongly increases up to a certain maximum and then again decreases for positive voltages to an even lower value. Two points are worth mentioning: first, the thickest sample reaches capacitance values similar to those at positive voltage values and, second, the thinnest sample does not reach so high values but returns to values similar to those exhibited by the as-deposited sample. Moreover, maximum capacitance occurs at positive voltages (0.3 V) for the 18.7 and 25.1 nm thick samples, and at a negative voltage (− 0.4 V) for the 7.7 nm thick sample. Current conduction measurements show parallel trends: the I–V curves for the same samples are plotted in Fig. 3. These curves were measured on fresh devices, i.e. in samples that were not previously biased. The
H. Castán et al. / Thin Solid Films 591 (2015) 55–59
57
Fig. 2. Capacitance–voltage curves measured at room temperature corresponding to HoTiOx-based MIM structures with Ho:Ti atomic ratios of 0.9, 1.4, and 1.5.
measurements were carefully carried out to avoid resistance switching: the voltage was first decreased from 0 to −1 V and then the current was recorded during its slow increment (sweep rate = 5 mV/s) up to +1 V. In all the cases, the annealed samples show higher current values than the as-deposited ones indicating some degradation of the insulating layer. In addition, the thinnest sample shows the poorest current behavior even before annealing. Holmium content also plays an important role in the linearity of capacitance of the as-deposited samples. Fig. 4(a) and (b) shows C–V curves for the as-deposited and annealed samples with the lowest values of holmium content (Ho:Ti = 0.54 and 0.48). These curves are all clearly non-linear, even for the as-deposited samples. In addition, in all cases the capacitance shows a maximum at negative values (about − 0.5 V) and reaches lower values at positive bias voltages. In Fig. 4(b) we see that the annealing process reduces the capacitance value, just in the opposite direction observed in the previous samples. As for the current values, one can see in Fig. 5 that these samples show poor fresh I–V curves and, once again, the sample with higher capacitance (as-deposited 23.8 nm-thick, Ho:Ti = 0.48) is the most leaky. These current values are too high for device application. On the other hand, it has been widely reported that defective dielectrics exhibit resistance switching effect that can be exploited to fabricate nonvolatile memory devices [6,7]. We have detected this behavior in the three samples with the lowest current density values (Fig. 3): for 18.7 (both as-deposited and annealed) and 25.1 nm-thick (as deposited) HoTiOx-based MIM samples, with Ho:Ti atomic ratio of 0.9 and 1.5, respectively, and we have also observed resistive switching in the
Fig. 3. Current density of pristine samples: HoTiOx-based MIM structures with Ho:Ti atomic ratios of 0.9, 1.4, and 1.5.
Fig. 4. Capacitance–voltage curves measured at room temperature corresponding to annealed HoTiOx-based MIM structures with Ho:Ti atomic ratios of 0.54 and 0.48 (a), and for the annealed and as-deposited samples with Ho:Ti atomic ratio of 0.48 (b).
sample with no titanium content (22.7 nm-thick stoichiometric Ho2O3). Resistive switching was not observed in samples with the lowest Ho:Ti ratio values, i.e. for 27.5 and 23.8 nm-thick HoTiOx based-MIM samples with Ho:Ti atomic ratio of 0.54 and 0.48, respectively. The thinnest sample (7.7 nm-thick HoTiOx sample with Ho:Ti atomic ratio of 1.4) does not show resistive switching behavior either. Therefore, both
Fig. 5. Current density of pristine samples: HoTiOx-based MIM structures with Ho:Ti atomic ratios of 0.54 and 0.48.
58
H. Castán et al. / Thin Solid Films 591 (2015) 55–59
Fig. 7. (a) Bipolar resistive switching cycles corresponding to the as-deposited 18.7 nm-thick HoTiOx with a Ho:Ti atomic ratio of 0.9 based MIM sample. (b) Endurance plot of current at 0.5 V positive and negative voltages.
Fig. 6. Three resistive-switching cycle plots for MIM samples of as-deposited 25.1 nm-thick HoTiOx with a Ho:Ti atomic ratio of 1.5 (a), and a pure 22.7 nm thick H2O3 annealed sample (b) films.
thickness and holmium content seem to be the key parameters to obtain useful devices. In Fig. 6(a) and (b), three resistive switching cycles for HoTiOx (25.1 nm-thick, Ho:Ti = 1.5) and Ho2O3 (22.7 nm-thick) samples, respectively, are plotted. In this figure one can see that the I–V cycles show bipolar resistive switching, since the bias voltage polarity has to be changed for achieving transitions between states. In the case of the 25.1 nm-thick layer, there are up to five orders of magnitude between the low resistance state (LRS) and the high resistance state (HRS) values, both measured at + 1 V. As the leakage current increases, the change in resistance decreases; in fact, the low resistivity current value is very similar for all the devices. In the Ho2O3 MIM sample, the LRS current measured at +1 V decreased whereas resistance reaches a value of around 60 MΩ, a really adequate value to achieve low power consumption. We have further investigated the RS behavior of our devices by performing 35 consecutive I–V cycles on the as-deposited 18.7 nm-thick HoTiOx-based MIM sample with Ho:Ti = 0.9 (see Fig. 7(a)). We observe larger current variability for HRS than for LRS, both at positive and negative voltages. This behavior can be more clearly observed in Fig. 7(b) in which the current values are plotted between +0.5 and −0.5 V for both states. This can be understood taken into account the size of the filamentary path and its susceptibility to local changes in the defects configurations. Very convenient resistance values are obtained for LRS (around 1 MΩ) and HRS (lower than 100 MΩ). However, the endurance is still poor: after the cycle number 30 the current at the
low resistance state gradually decays down to the HRS. Therefore, although these samples are promising and do show RS effect, more work has to be made in order to make them good challengers for RRAM device fabrication. To conclude, a remarkable feature exhibited by the post-breakdown HoTiOx I–V characteristic is illustrated in Fig. 8. This figure shows that the curves can be nicely fitted using a power-law model I = aVb (Fig. 8) in a very wide voltage and current range. This behavior is typical of the soft breakdown failure mode (HRS) in ultrathin (b5 nm) SiO2 layers [20] but has also been reported for a variety of high-k dielectrics [21–23]. In the present case, the soft breakdown mode occurs in a
Fig. 8. Experimental (colored lines) and modeled (black lines) I–V curves corresponding to the annealed 18.7 nm-thick HoTiOx with a Ho:Ti atomic ratio of 0.9 based MIM sample. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
H. Castán et al. / Thin Solid Films 591 (2015) 55–59
relatively thick (around 18.7 nm) film. Earlier reports indicate that RS has been observed in even thicker TiO2 films [24]. Interestingly, in HoTiOx the power-law model can be applied up to the hard breakdown mode (LRS) with a small deviation at the largest current values in the positive bias range because of the potential drop in the semiconductor substrate (inversion condition). 4. Conclusions Atomic layer deposited films of holmium titanium oxide (HoTiOx) have been studied in order to explore their suitability to be exploited for conventional gate dielectric material and nonvolatile memory applications. HoTiOx MIM structures clearly exhibit resistive switching effect. However, the number of LRS to HRS switching cycles is still low and future work must be done in order to improve the endurance of these promising structures. It was also shown that the post-breakdown I–V characteristics can be fitted using a power-law model both in the LRS and HRS regimes. This extends the validity of this model to the hard breakdown regime. Acknowledgments The study has been supported by the Spanish TEC2014 under grant No. 52152-C3-3-R, and by the Finnish Centre of Excellence in Atomic Layer Deposition. E. Miranda acknowledges the Ministerio de Economía y Competitividad, Spain (TEC 2012-32305) with support of federal funds. References [1] T.M. Pan, M.D. Huang, Structural properties and sensing characteristics of high-k Ho2O3 sensing film-based electrolyte–insulator–semiconductor, Mater. Chem. Phys. 129 (2011) 919–924. [2] G. Darbandy, F. Lime, A. Cerdeira, M. Estrada, I. Garduño, B. Iñiguez, Study of potential high-k dielectric for UTB SOI MOSFETS using analytical modeling of the gate tunneling leakage, Semicond. Sci. Technol. 26 (2011) 115002. [3] T.M. Pan, M.D. Huang, C.W. Lin, M.H. Wu, Development of high-k HoTiO3 sensing membrane for pH detection and glucose biosensing, Sensors Actuators B Chem. 144 (2010) 139–145. [4] M. Seo, S.H. Rha, S.K. Kim, J.H. Han, W. Lee, S. Han, C.S. Hwang, The mechanism for the suppression of leakage current in high dielectric TiO2 thin films by adopting ultra-thin HfO2 films for memory application, J. Appl. Phys. 110 (2011) 024105. [5] S.K. Kim, S.W. Lee, J.H. Han, B. Lee, S. Han, C.S. Hwang, Capacitors with an equivalent oxide thickness of b0.5 nm for nanoscale electronic semiconductor memory, Adv. Funct. Mater. 20 (2010) 2989–3003. [6] R. Waser, M. Aono, Nanoionics-based resistive switching memories, Nat. Mater. 6 (2007) 833–840.
59
[7] D.S. Jeong, R. Thomas, R.S. Katiyar, J.F. Scott, H. Kohlstedt, A. Petraru, C.S. Hwang, Emerging memories: resistive switching mechanisms and current status, Rep. Prog. Phys. 75 (2012) 076502. [8] C. Rohde, B.J. Choi, D.S. Jeong, S. Choi, J.S. Zhao, C.S. Hwang, Identification of a determining parameter for resistive switching of TiO2 thin films, Appl. Phys. Lett. 86 (2005) 262907. [9] D. Lee, H. Choi, H. Sim, D. Choi, H. Hwang, M.J. Lee, S.A. Seo, I.K. Yoo, Resistance switching of the nonstoichiometric zirconium oxide for nonvolatile memory applications, IEEE Electron Device Lett. 26 (2005) 719–721. [10] S. Spiga, A. Lamperti, C. Wiemer, M. Perego, E. Cianci, G. Tallarida, H.L. Lu, M. Alia, F.G. Volpe, M. Fanciulli, Resistance switching in amorphous and crystalline binary oxides grown by electron beam evaporation and atomic layer deposition, Microelectron. Eng. 85 (2008) 2414–2419. [11] Y. Li, S. Long, M. Zhang, Q. Liu, L. Shao, S. Zhang, Y. Wang, Q. Zuo, S. Liu, M. Liu, Resistive switching properties of Au/ZrO2/Ag structure for low-voltage nonvolatile memory applications, IEEE Electron Device Lett. 31 (2010) 117–119. [12] D. Acharyya, A. Hazra, P. Bhattacharyya, A journey towards reliability improvement of TiO2 based resistive random acces memory: a review, Microelectron. Reliab. 54 (2014) 541–560. [13] B. Hudec, A. Paskaleva, P. Jancovic, J. Dérer, J. Fedor, A. Rosová, E. Dobrocka, K. Fröhlich, Resistive switching in TiO2-based metal–insulator–metal structures with Al2O3 barrier layer at the metal/dielectric interface, Thin Solid Films 563 (2014) 10–14. [14] K.M. Kim, B.J. Choi, C.S. Hwang, Localized switching mechanism in resistive switching of atomic layer deposited TiO2 thin films, Appl. Phys. Lett. 90 (2007) 242906. [15] K.M. Kim, B.J. Choi, Y.C. Shin, S. Choi, C.S. Hwang, Anode-interface localized filamentary mechanism in resistive switching of TiO2 thin films, Appl. Phys. Lett. 91 (2007) 012907. [16] K.M. Kim, T.H. Park, C.S. Hwang, Dual conical conducting filament model in resistance switching TiO2 thin films, Sci. Rep. 5 (2015) 7844, http://dx.doi.org/10.1038/ srep07844. [17] T. Suntola, Atomic layer epitaxy, Thin Solid Films 216 (1992) 84–89. [18] K. Kukli, M. Kemell, M.C. Dimri, E. Puukilainen, A. Tamm, R. Stern, M. Ritala, M. Leskelä, Holmium titanium oxide thin films grown by atomic layer deposition, Thin Solid Films 565 (2014) 261–266. [19] H. García, H. Castán, S. Dueñas, E. Pérez, L.A. Bailón, K. Kukli, M. Ritala, M. Leskelä, Resistive switching behavior and electrical properties of TiO2:Ho2O3 and HoTiOx based MIM capacitors, Mater. Res. Soc. Symp. Proc. 1691 (2014), http://dx.doi.org/ 10.1557/opl.2014.598. [20] E. Miranda, J. Suñé, R. Rodríguez, M. Nafría, X. Aymerich, A function-fit model for the soft breakdown failure mode, IEEE Electron Device Lett. 20 (1999) 265–267. [21] M. Houssa, P.W. Mertens, M.M. Heyns, J.S. Jeon, A. Halliyal, B. Ogle, Soft breakdown in very thin Ta2O5 gate dielectric layers, Solid State Electron. 44 (2000) 521–525. [22] E. Miranda, J. Martin-Martinez, E. O'Connor, G. Hughes, P. Casey, K. Cherkaoui, S. Monaghan, R. Long, D. O'Connell, P.K. Hurley, Effects of the electrical stress on the conduction characteristics of metal gate/MgO/InP stacks, Microelectron. Reliab. 49 (2009) 1052–1055. [23] M.B. González, J.M. Rafí, O. Beldarrain, M. Zabala, F. Campabadal, Charge trapping and electrical degradation in atomic layer deposited Al2O3 films, Microelectron. Eng. 109 (2013) 57–59. [24] B. Hudec, M. Hranai, K. Hušekova, J. Aarik, A. Tarre, K. Fröhlich, RuO2/TiO2 based MIM capacitors for DRAM application, The 8th Int. Conf. on Advanc. Semicond. Dev. and Microsystems (ASDAM 2010), Smolenice, Slovakia, 25–27 Oct. 2010, IEEE, Piscataway, NJ, USA 2010, pp. 255–258.