J. Niinistö, M. Ritala, and M. Leskelä Department of Chemistry, University of Helsinki, Helsinki, Finland

Transcription

1 Electrical properties of thin zirconium and hafnium oxide high-k gate dielectrics grown by atomic layer deposition from cyclopentadienyl and ozone precursors S. Dueñas, a H. Castán, H. Garcia, A. Gómez, and L. Bailón Departamento de Electricidad y Electrónica, E. T. S. I. Telecomunicación, Universidad de Valladolid, Valladolid, Spain K. Kukli Institute of Physics, Department of Materials Science, University of Tartu, Tartu, Estonia and Department of Chemistry, University of Helsinki, Helsinki, Finland J. Niinistö, M. Ritala, and M. Leskelä Department of Chemistry, University of Helsinki, Helsinki, Finland Received 28 May 2008; accepted 22 September 2008; published 9 February 2009 ZrO 2 and reference HfO 2 films grown by atomic layer deposition from metal cyclopentadienyls and ozone as precursors to thicknesses ranging from 3.6 to 13.1 nm on etched silicon showed electrical characteristics adequate to high-k dielectrics. The best results in terms of low interface state densities were obtained when CpMe 2 ZrMe 2 precursor was used, with Cp denoting the cyclopentadienyl group C 5 H 5, and Me the methyl group CH 3. The ZrO 2 films grown from CpMe 2 Zr OMe Me possessed nearly an order of magnitude higher trap state densities. Similar dependence on the precursor chemistry was observed upon recording the flatband voltage time transients. The flatband voltage transients, originating from phonon-assisted tunneling between localized states at oxide silicon interface, were the lowest in HfO 2 films grown from CpMe 2 Hf OMe Me. The leakage current densities were also lower in the HfO 2 films, compared to ZrO 2. On the other hand, interfacial trap state densities in HfO 2 based capacitors remained higher than those measured in the case of ZrO 2 films. Process-dependent qualities of the capacitors have been described. At the same time, the current conduction mechanisms in all films were essentially bulk driven, not affected noticeably by the interfacial barriers American Vacuum Society. DOI: / I. INTRODUCTION ZrO 2 and HfO 2 have gained considerable attention in the microelectronic industry in order to replace SiO 2 as capacitor dielectric material, 1 at first because of their high permittivity, but also due to large band gap, related high breakdown field, and sufficient thermodynamic stability on Si. Although the interfacial stability of HfO 2 on Si tends to be superior to that of ZrO 2, 2 ZrO 2 is still an intriguing material due to its greater ability to form higher-permittivity metastable crystallographic polymorphs. Atomic layer deposition ALD of ZrO 2 and HfO 2 films has been realized using different precursors, most often or traditionally from chlorides, 3 5 but also from alternative, e.g., organometallic cyclopentadienyl-based precursors Good-quality ZrO 2 and HfO 2 films can be grown using ALD processes established for CpMe 2 ZrMe 2, 9,11 CpMe 2 Zr OMe Me, 9,11 CpMe 2 Hf OMe Me, 9,10 and O 3 as precursors. CpMe 2 ZrMe 2 may allow better stability of the growth rate against temperature parametrization, 11 whereas the CpMe 2 Hf OMe Me tended to allow relatively higher growth rates 11 and, possibly, stronger contribution from cubic/tetragonal phases in the films grown. 10 At the same time, also lower leakage was achieved in the ZrO 2 grown a Electronic mail: sduenas@ele.uva.es from CpMe 2 Zr OMe Me. 9 The lowest leakage currents have still been characteristic of HfO 2 films. 9 All these precursors have demonstrated comparable abilities to provide conformal growth over three-dimensional substrates suited to the development of nanoelectronic devices. 9,10 In order to further evaluate the potential electrical stability of the films grown in such processes likely exhibiting characteristic advantages and drawbacks, more detailed studies are required. In this work, we have investigated the defect densities and their effect to the capacitive properties and conduction mechanisms of metal-oxide-semiconductor MOS structures with ZrO 2 and, for comparison, some HfO 2 layers atomic layer deposited by using the precursors mentioned above. The electrical characterization techniques used were capacitance-voltage C-V, deep-level transient spectroscopy DLTS to obtain the interfacial state density values, conductance transients G-t to obtain the disordered-induced gap states DIGS densities, current-voltage I-V, and constantcapacitance flatband voltage transients V FB -t. II. EXPERIMENT A. Sample preparation The measured samples were high-k dielectric based MOS structures. The metal oxide films were grown on HF-etched 389 J. Vac. Sci. Technol. B 27 1, Jan/Feb /2009/27 1 /389/5/$ American Vacuum Society 389

2 390 Dueñas et al.: Electrical properties of thin zirconium oxide 390 TABLE I. Growth parameters and thickness of the atomic layer deposited ZrO 2 and HfO 2 films. Cp denotes the cyclopentadienyl group C 5 H 5,Me denotes the methyl group CH 3. Precursor p-type silicon substrates by ALD in a hot-wall flow-type reactor F The metal precursors, amounts of growth cycles and growth temperatures are listed in Table I. The growth temperatures were chosen considering the optimum temperature range for a particular precursor applied. The cycle times used were s for the sequence metal precursor pulse-purge-o 3 pulse-purge. The thicknesses of the films, also shown in Table I, were evaluated from x-ray reflection patterns. The electrical measurements were carried out on Al/ZrO 2 / p-si 100 /Al or Al/HfO 2 / p-si 100 /Al capacitors. Aluminum dot electrodes with an area of mm 2 were e-beam evaporated on top of the dielectric layers through a shadow mask. To form nearly Ohmic contacts to silicon substrates, the back sides of the wafers were etched in HF and metallized by evaporating a 100 nm thick Al layer. B. Measurement procedure Cycles at T growth at 350 C Thickness by XRR nm CpMe 2 ZrMe CpMe 2 ZrMe CpMe 2 ZrMe CpMe 2 Zr OMe Me CpMe 2 Zr OMe Me CpMe 2 Zr OMe Me CpMe 2 Hf OMe Me CpMe 2 Hf OMe Me In order to record the electrical parameters at several temperatures varying between 77 K and room temperature, the samples were first cooled in darkness from room temperature to 77 K at zero bias in an Oxford DM1710 cryostat. An Oxford ITC 502 controller was used to monitor the temperature during the measurements. The C-V measurements were carried out with the assistance of a Boonton 72B capacitance meter and a Keithley 617 programmable electrometer. The capacimeter makes measurements of shunt capacitance by applying a 1 MHz and 15 mv level ac signal test. Voltage bias consists of 50 mv steps. The capacitance measurements are made 10 s after the bias-voltage step is applied in order to prevent transient instabilities in the capacitance values. Interfacial state densities were obtained by saturating pulse DLTS. The bias voltage was chosen for each sample so that the capacitor was just at the limit between depletion and weak inversion, and a 10 ms wide pulse high enough to drive the capacitors into accumulation was applied in order to fill all interface traps. Information on the traps was obtained by analyzing the capacitance transient that results as the traps empty, i.e., return to equilibrium. The experimental setup of the conductance transient technique consist of an HP 33120A arbitrary wave-form generator to apply the bias pulses, an EG&G 5206 two-phase lock-in analyzer to measure the conductance, and an HP 54501A digital oscilloscope to record the complete conductance transient. The I-V curves were measured with a Keithley 6517A programmable electrometer in the stair sweep voltage mode while the voltage step used was 5 mv. Finally, to obtain the flatband voltage transients we have implemented a feedback system that varies the applied gate voltage accordingly to keep constant the flatband capacitance value. The capacitance is measured by using a Boonton 72B capacimeter, and an electrometer Keithley 6517A is used to read the capacimeter analog output as well as to send the data to the computer. The modular source Agilent N6700 working with the N6761A module provides the bias voltage. III. RESULTS AND DISCUSSION A. Structure and composition of the films Structural studies have indicated earlier that ZrO 2 films grown from CpMe 2 Zr OMe Me contained only metastable tetragonal or cubic polymorph. 9 In the ZrO 2 films grown from CpMe 2 ZrMe 2 and HfO 2 grown from CpMe 2 Hf OMe Me, also monoclinic polymorphs became apparent when the ZrO 2 films reached a thickness of 12 nm and HfO 2 films 8 nm. 9 The films thinner than 4 5 nm remained amorphous. The crystallized films consisted of dense homogeneous grains, although with distinct grain boundaries. Approximately nm thick interfacial layer was formed between the film and substrate. The interfacial layer is probably silicon oxide, formed due to the strong oxidizing capability of ozone used as the oxygen precursor. Elastic recoil detection analysis has earlier verified the stoichiometry of HfO 2 grown from CpMe 2 Hf OMe Me and O 3, while the concentration of residual carbon and hydrogen remained lower than 0.1 at. %. 10 Rutherford backscattering spectroscopy studies have earlier been carried out on ZrO 2, demonstrating low levels of carbon residuals below the detection limit, 1 at. %. 11 Hydrogen content could not be analyzed. However, the oxygen to zirconium atomic ratio could reach in the films grown from CpMe 2 Zr OMe Me at 300 C and in the films grown from CpMe 2 ZrMe 2 at 350 C, 11 referring to the possible presence of hydroxyl groups or carbonates, slightly increasing the relative content of oxygen. B. Electrical characterization results Normalized 1 MHz C-V curves measured at 77 K for ZrO 2 and HfO 2 film-based MOS devices are shown in Fig. 1. C-V measurements have also been carried out at room temperature, but the higher leakage currents yield more distorted and confusing curves than those obtained at low temperatures. Since the flatband voltage V FB value results mainly from the difference in work function values between the Al electrode and Si substrate ms 0.5 V and the presence of fixed charges, flatband voltage shift is related to the presence of charge in the dielectric films. It is apparent that when CpMe 2 ZrMe 2 precursor is used, flatband voltage displacements are lower than in the case of CpMe 2 Zr OMe Me and J. Vac. Sci. Technol. B, Vol. 27, No. 1, Jan/Feb 2009

3 391 Dueñas et al.: Electrical properties of thin zirconium oxide 391 FIG. 1. Normalized 1 MHz C-V curves measured at 77 K for ZrO 2 and HfO 2 film-based MOS devices. CpMe 2 Hf OMe Me precursors. In the last case, we can also observe that hafnium oxide based samples exhibit less flatband voltage displacements than zirconium oxide based samples. On the other hand, the thinnest films show considerable amount of hysteresis in comparison to thicker samples. Finally, the 3.6 nm thick sample, processed by using CpMe 2 ZrMe 2, exhibited a flatband voltage displacement around 1 V less than the other samples, so it seems that the CpMe 2 ZrMe 2 precursor leads to better quality thin films. This last result must be taken by caution due that it has been only observed for one sample. From the flatband voltage shift of C-V curves, values between Ccm 2 and cm 2 of the fixed oxide charge densities are estimated. DLTS measurements Fig. 2 indicated that HfO 2 films possessed higher interfacial state densities D it than ZrO 2 samples. Also, ZrO 2 films grown from the CpMe 2 ZrMe 2 precursor had lower interface trap densities than the films grown from FIG. 3. Current-electric field dependency fitting following the Poole-Frenkel model: at different temperatures corresponding to an 8.3 nm thick hafnium oxide based MOS sample a, and at room temperature corresponding to several hafnium oxide and zirconium oxide based MOS samples b. FIG. 2. Interfacial state density distributions obtained by DLTS for ZrO 2 and HfO 2 film-based MOS devices. CpMe 2 Zr OMe Me. For both HfO 2 and ZrO 2, D it increases with film thickness. In good agreement with C-V measurements, DLTS results indicate that the best quality interface corresponds to the 3.6 nm thick sample processed by using CpMe 2 ZrMe 2. The conductance transients recorded were as small as G/ Hz 1 at temperatures between 200 and 300 K, and undetectable for our experimental setup at temperatures below 200 K, corresponding to DIGS densities lower than cm 2 ev 1. Prebreakdown current densities were lower in HfO 2 films, compared to ZrO 2. Current-voltage characteristics of all the samples were well fitted according to the Poole-Frenkel PF currents Fig. 3, indicating that the main conduction mechanism is bulk related. 13 Figure 3 a shows the plot of I/E in logarithmic scale against E 1/2 at several temperatures, corresponding to the 8.3 nm thick hafnium oxide based MOS sample. The PF field-lowering coefficient values, PF, are JVST B-Microelectronics and Nanometer Structures

4 392 Dueñas et al.: Electrical properties of thin zirconium oxide 392 FIG. 4. Normalized flatband transients measured at room temperature corresponding to several hafnium oxide and zirconium oxide based MOS samples. FIG. 5. Flatband transients measured at different temperatures, and the Arrhenius plot corresponding to a 3.6 nm thick zirconium oxide based MOS sample a, and to an 8.3 nm thick hafnium oxide based MOS sample b. about ev m 1/2 V 1/2. The theoretical value of PF parameter obtained from the dielectric constant of HfO 2 layer at optical frequencies about 3.6 is ev m 1/2 V 1/2, so the experimental and theoretical values do not completely agree. The good fit of the Poole- Frenkel plots can indicate that this difference is related to the physical structure of the films, especially to the existence of an interfacial layer. In Fig. 3 b the PF fits at room temperature corresponding to all samples are depicted. When the CpMe 2 M OMe Me M =Zr,Hf precursors are used, PF values of about ev m 1/2 V 1/2 are obtained. However, the values of PF are lower around ev m 1/2 V 1/2 in the cases of CpMe 2 ZrMe 2 precursor. Flatband voltage transients recorded at different temperatures provide valuable information about phonon-assisted tunneling mechanisms These transients are recorded under conditions without external stress, and so they are originated by phonon-assisted tunneling between localized states, phonons producing the ionization of traps existing in the bandgap of the insulator. Figure 4 shows normalized flatband voltage transients measured at room temperature for the hafnium oxide and zirconium oxide based MOS samples. Transient amplitudes clearly depend on the precursor and the dielectric film. In fact, the transient amplitudes of Al/ZrO 2 / p-si samples fabricated by using the CpMe 2 Zr OMe Me precursor are greater than those obtained when the CpMe 2 ZrMe 2 precursor is used. Also, flatband voltage transients of Al/HfO 2 / p-si samples exhibit lower amplitude than those corresponding to Al/ZrO 2 / p-si samples, in good agreement with the current-voltage behavior. As it has been shown elsewhere, 14 the amplitude of flatband voltage transients also depends on temperature. Indeed, their magnitude follows an Arrhenius plot that provides the activation energy of the soft-optical phonons involved in the tunneling mechanism. As it is shown in Fig. 5, activation energy values of about mev are obtained. Fouriertransform infrared measurements could give straightforward evidence that such mechanism occurs indeed. One can follow systematic differences between the films grown from different precursors, especially in terms of interface state densities, D it Fig. 2. The ZrO 2 based capacitors were less defective in terms of D it, compared to HfO 2, and, within the ZrO 2 serial, CpMe 2 ZrMe 2 occurred more advantageous compared to CpMe 2 Zr OMe Me. Also the flatband transients have been weaker in the films grown from CpMe 2 ZrMe 2. The chemical purity, i.e., the residual contamination of ZrO 2 has not been lower compared to HfO 2, but rather slightly excessive oxygen has earlier been considered. Therefore, the electronic defect density and stability might rather be connected to the crystallographic phase composition, at least partially. Indeed, under the deposition conditions applied, stoichiometric stable monoclinic polymorph tended to appear at somewhat lower thicknesses in ZrO 2 deposited from CpMe 2 ZrMe 2, compared to the films grown from CpMe 2 Zr OMe Me. 9 It is thus possible that the crystallographic phase heterogeneity may not always result in greater density of electronic defects affecting the capacitive J. Vac. Sci. Technol. B, Vol. 27, No. 1, Jan/Feb 2009

5 393 Dueñas et al.: Electrical properties of thin zirconium oxide 393 behavior of MOS devices, provided that only stable stoichiometric phases can be formed at early stages of growth. This, however, does not explain the even greater density of defects observed in HfO 2 -based MOS structures. Besides, the D it increases with film thickness in general and thus with crystallographic phase heterogeneity within each series grown using particular growth chemistry Fig. 2. Nevertheless, although the behavior of D it and the extent of flatband voltage transients cannot yet be fully explained on the basis of film purity and crystallinity, the electronic behavior recorded carries the features evidently characteristic of the process applied, including precursor chemistry and deposition temperature. IV. CONCLUSIONS The electrical characterization of ALD ZrO 2 and HfO 2 films as deposited by using cyclopentadienyl based precursors CpMe 2 Zr OMe Me, CpMe 2 ZrMe 2, and CpMe 2 Hf OMe Me, with O 3, has demonstrated the suitability of these precursors for the fabrication of high-k dielectric materials. In terms of interface trap densities, ZrO 2 films grown to thickness lower than 4 nm using CpMe 2 ZrMe 2 occurred most promising, showing D it values cm 2 ev 1, whereas in the films grown from CpMe 2 Zr OMe Me the D it values exceeded cm 2 ev 1.InHfO 2 films, at the same time, the D it values reached cm 2 ev 1 and higher. Flatband voltage transients, i.e., instabilities of C-V curves, were also weaker in ZrO 2 films grown from CpMe 2 ZrMe 2, compared to the films grown from CpMe 2 Zr OMe Me. The flatband transients were even weaker in HfO 2 films, compared to the ZrO 2 films. Also the prebreakdown current densities were lower in HfO 2, indicating, expectably, better insulating properties compared to ZrO 2 despite the higher interfacial trap densities. The electrical quality of the high-k films has thus been rather strongly dependent on the material deposited as well as the precursor chemistry applied. In all the films, however, the currents were driven by the bulk properties of the films, rather than interfaces between electrodes and oxide films, as revealed by the Poole-Frenkel characteristics. ACKNOWLEDGMENTS The study was partially supported by the local government Junta de Castilla y León under Grant No. VA018A06, and by the Spanish TEC2005 under Grant No /MIC. The authors wish to thank M. Putkonen for depositing part of the thin film samples. 1 G. D. Wilk, R. M. Wallace, and J. M. Anthony, J. Appl. Phys. 89, M. Gutowski, J. E. Jaffe, C. L. Liu, M. Stoker, R. I. Hegde, R. S. Rai, and P. J. Tobin, Appl. Phys. Lett. 80, M. Ritala and M. Leskelä, Appl. Surf. Sci. 75, M. Copel, M. Gribelyuk, and E. Gusev, Appl. Phys. Lett. 76, J. Aarik, A. Aidla, H. Mändar, T. Uustare, and V. Sammelselg,Thin Solid Films 408, M. Putkonen and L. Niinistö, J. Mater. Chem. 11, M. Putkonen, J. Niinistö, K. Kukli, T. Sajavaara, M. Karppinen, H. Yamauchi, and L. Niinistö, Chem. Vap. Deposition 9, J. Niinistö, M. Putkonen, L. Niinistö, K. Kukli, M. Ritala, and M. Leskelä, J. Appl. Phys. 95, K. Kukli et al., Microelectron. Eng. 84, J. Niinistö, M. Putkonen, L. Niinistö, F. Song, P. Williams, P. N. Heys, and R. Odedra, Chem. Mater. 19, J. Niinistö et al., J. Mater. Chem. 18, T. Suntola, Thin Solid Films 216, J. Frenkel, Phys. Rev. 54, S. Dueñas, H. Castán, H. García, L. Bailón, K. Kukli, T. Hatanpää, M. Ritala, and M. Leskelä, Microelectron. Eng. 47, S. Dueñas et al., J. Electrochem. Soc. 154, G S. Dueñas, H. Castán, H. García, A. Gómez, L. Bailón, M. Toledano- Luque, I. Mártil, and G. González-Díaz, Semicond. Sci. Technol. 22, JVST B-Microelectronics and Nanometer Structures