- Email: [email protected]
Annealing effect on physical and electrical characteristics of thin HfO2, HfSixOy and HfOyNz films on Si
Microelectronic Engineering 86 (2009) 357–360
Contents lists available at ScienceDirect
Microelectronic Engineering journal homepage: www.elsevier.com/locate/mee
Annealing effect on physical and electrical characteristics of thin HfO2, HfSixOy and HfOyNz films on Si Joo-Hyung Kim a,c,*, Velislava A. Ignatova a, Martin Weisheit b a
Fraunhofer Institute, Center of Nanoelectronic Technologies (CNT), Königsbrücker Str., 01099 Dresden, Germany Center for Complex Analysis, AMD Fab36 LLC & Co. KG, D-01109 Dresden, Germany c Dept. of Mechanical Engineering, Inha University, 253 Young-Hyun Dong, Nam Gu, Incheon, Republic of Korea b
a r t i c l e
i n f o
Article history: Received 13 June 2008 Received in revised form 30 October 2008 Accepted 7 November 2008 Available online 24 November 2008 Keywords: HfO2 Gate oxide High-k Annealing Leakage current Bandgap
a b s t r a c t We report the effect of annealing on electrical and physical characteristics of HfO2, HfSixOy and HfOyNz gate oxide films on Si. Having the largest thickness change of 0.3 nm after post deposition annealing (PDA), HfOyNz shows the lowest leakage current. It was found for both as-grown and annealed structures that Poole–Frenkel conduction is dominant at low field while Fowler–Nordheim tunneling in high field. Spectroscopic ellipsometry measurement revealed that the PDA process decreases the bandgap of the dielectric layers. We found that a decreasing of peak intensity in the middle HfOyNz layer as measured by Tof-SIMS may suggest the movement of N toward the interface region between the HfOyNz layer and the Si substrate during the annealing process. Crown Copyright Ó 2008 Published by Elsevier B.V. All rights reserved.
1. Introduction Aggressive downscaling of CMOS circuits produces high leakage current due to direct tunneling of electrons through the gate SiO2 layer (thickness <1.5 nm). According to the International Technology Roadmap for Semiconductors (ITRS), to keep the equivalent oxide thickness with low leakage current in the gate oxide, alternative high dielectric constant (high-k) oxides such as ZrO2, HfO2 and Ta2O5 have been considered as possible candidates for further downscaling [1–4]. As the best candidate for gate oxide application, HfO2 has a high dielectric constant (ranging from 20 to 49 for the monoclinic, cubic and tetragonal phase [5–7], a large band gap with sufficient band offset (>1.5 eV) and is thermally stable in contact with silicon substrates [8,9]. The main drawback of HfO2 is the low crystallization temperature (T < 400 °C) which can promote crystal boundaries within the layers to act as leakage paths and impurity getters during processing. Therefore it is an important issue to increase the kinetic stability while keeping the advantages of HfO2 for its adoption into semiconductor manufacturing. The first method to increase the crystallization temperature of HfO2 is to form a silicate of HfO2, * Corresponding author. Address: Dept. of Mechanical Engineering, Inha University, 253 Young-Hyun Dong, Nam Gu, Incheon, Republic of Korea. Tel.: +82 32 874 7325; fax: +82 32 832 7325. E-mail addresses: [email protected], [email protected] (J.-H. Kim).
but then the dielectric constant decreases [10]. The other option is to form an aluminate layer of HfO2 which has a rather high dielectric constant due to Al2O3 itself being a high-k material [11]. Up to date, however, few publications show the annealing effect on the electrical performance of the dielectric layer and especially the elemental depth profile in the layer. This report presents the effect of annealing on the electrical and physical characteristics of Hf-based gated oxides. 2. Experimental procedure HfO2, HfSixOy and HfOyNz films with thicknesses around 4 nm were deposited on HF cleaned 1200 n-type Si substrates by atomic layer deposition (ALD). Tetrakis(ethylmethylamino)-hafnium (TEMA-Hf) and Tetrakis(ethylmethylamino)-silicon (TEMA-Si) plus ozone were used as precursors. Ion implantation, reported to be successfully integrated into a traditional CMOS process of N2+ in ALD was introduced to form the HfOyNz layer. After deposition, in order to investigate the annealing effect on Hf-based gate oxides, one group of samples was subsequently annealed at 1000 °C, 10 s in N2 ambient. In order to facilitate electrical characterization, top electrodes made up of 400 nm thick Al dots with different diameters – 350, 600 and 1300 lm – were deposited by sputtering on both the as-grown and annealed samples. The electrical properties were characterized by a semiconductor parameter analyser (HP 4156 A). Capacitance–voltage measurement was performed
0167-9317/$ - see front matter Crown Copyright Ó 2008 Published by Elsevier B.V. All rights reserved. doi:10.1016/j.mee.2008.11.012
358
J.-H. Kim et al. / Microelectronic Engineering 86 (2009) 357–360
by a precision LCR meter (HP 4284A) with parallel mode (Cp) at 1 kHz. Structural investigation was done by Transmission Electron Microscopy (TEM, FEI TECNAI F30). To obtain detailed information about the annealing effect on the dielectric layer as well as the interfaces between layers, we used time-of-flight secondary ion mass spectrometry (ToF-SIMS, ION-TOF-V) with a 1 keV Cs+ primary ion beam at incident angle of 45° for elemental depth distribution and the interface composition. 3. Results and discussion The leakage current plot as a function of applied voltage of as-grown samples is shown in Fig. 1. The leakage current was obtained under electron injection from the Si substrate. For all asgrown samples the leakage current level is similar up to an applied voltage of 2.5 V while it becomes clearly different as the applied field is higher than 2.5 V. The highest leakage current was observed for the HfO2 sample in the whole range of applied fields. The HfSixOy and HfOyNz samples show a lower leakage current, which indicates Si and N incorporation into HfO2 has a role to decrease leakage current due to their high conduction band (CB) offset of 3.2 and 2.4 eV for SiO2 and Si3N4, respectively. Robertson [12] reports that the CB offset of HfSiO4 is 1.8 eV, which is between that of HfO2 (1.4 eV) and SiO2. The inset in Fig. 1 shows the two regions of different leakage current mechanisms of the HfOyNz sample: Trap assisted tunneling (TAT) and Poole–Frenkel (P–F) conduction in low electric field were observed [13] while Fowler–Nordheim (F–N) tunneling was dominant at high field. From the P–F plots in ln (J/E) vs. E1/2, the trap depth /PF of as-grown HfO2, HfSixOy and HfOyNz samples was 0.57 eV, 0.61 eV and 0.59 eV, respectively. It known that Si or N incorporation into HfO2 increases the trap depth of dielectric layers [5]. Fig. 2 presents the annealing effect on leakage current of Hfbased oxide samples. The leakage current of all annealed samples at 1 V is more than first order of magnitude higher than that of the as-grown samples. The leakage current mechanism is similar to the as-grown samples in Fig. 1, while the transition point from P–F to F–N occurs at higher field. Since the leakage current is related to the physical thickness and the band gap (BG) of a layer, we measured the thickness change and bandgap of as-grown and annealed samples by variable angle spectroscopic ellipsometry (Woollan VASE M-2000, spectral range 200–1000 nm). The measured data is summarized in Table 1. It is noteworthy that the biggest annealing effect on thickness
Fig. 1. Leakage current plot of as-grown HfO2 based metal–insulator–semiconductor (MIS) structures. The inset presents the P–F plot of the HfOyNz sample.
Fig. 2. Annealing effect on leakage current of HfO2 based MIS capacitors. The inset shows the region of Poole–Frenkel conduction (<1.0 V) and Fowler–Nordheim tunneling in high field range (>1.1 V). The crossover point from P–F to F–N is lower than for as-grown samples.
Table 1 Annealing effect on physical thickness of Hf-based gate oxides measured by spectroscopic ellipsometry.
As-grown Annealed
HfO2 (nm)
HfSixOy (nm)
HfOyNz (nm)
4.35 ± 0.02 4.25 ± 0.03
4.36 ± 0.02 4.17 ± 0.03
4.45 ± 0.02 4.15 ± 0.03
change was found in the HfOyNz sample. The dielectric functions of the films, which show a steep rise of absorption in the UV part of the spectrum, were parameterized by a Tauc-Lorentz-function. The bandgaps obtained from a fit of this parameterization to the ellipsometric data are 5.35, 5.40 and 5.38 eV for the as-grown HfO2, HfSixOy and HfOyNz, respectively. Regarding the annealing effect on the bandgap, shown in Fig. 3, it is clearly revealed that the annealing process promotes the decrease of the bandgap for all samples in this study. Wang et al reported that the formation of Si-rich Hf-silicate stems from the linkage between SiO2 and Hf precursor during the first HfO2 monolayer formation [14]. Since Hf-silicate is thermally stable in contact with Si and HfO2, silicate formation may occur during the deposition of HfO2 and HfOyNz. Additional annealing would promote silicate formation in the interface layer. To elucidate this point further, we characterized
Fig. 3. Bandgap changes of as-grown and annealed HfO2, HfSixOy and HfOyNz samples using variable angle spectroscopic ellipsometry (VASE).
J.-H. Kim et al. / Microelectronic Engineering 86 (2009) 357–360
359
Fig. 6. Annealing effect on carbon and fluorine contents of HfOyNz/SiO2/Si.
Fig. 4. Cross-sectional TEM image of the annealed HfOyNz/Si sample. The inset is a detail image of the interface between HfOyNz and Si.
the cross-section of the annealed HfOyNz sample by TEM. Fig. 4 shows that there is no amorphous layer formed at the interface, while the amorphous SiO2 layer before the dielectric layer deposition was around 1 nm. This indicates that the interfacial SiO2 layer seems to be decomposed during the dielectric layer deposition and annealing process [12]. The equivalent oxide thickness (EOT = thigh-kkSiO2/khigh-k, where thigh-k is the physical thickness of high-k layer, kSiO2 is the dielectric constant of SiO2, and khigh-k is the dielectric constant of high-k layer from capacitance–voltage measurement) was calculated from C–V measurements. The EOT values of as-grown Hf-based samples were 0.4–0.7 nm while those of the annealed samples were slight larger at around 0.5–0.6 nm. The dielectric constants were 31, 30 and 28 for the annealed HfO2, HfSixOy and HfOyNz, respectively, which are slightly higher than those of as-grown samples except HfSixOy. To get more detailed information about the chemical composition, we measured the elemental depth profile of the as-grown and annealed HfOyNz samples using secondary ion mass spectroscopy (SIMS). Fig. 5 shows the difference in the depth profiles of HfOyNz in as-grown and annealed samples. It was observed that the HfON secondary ion (atomic weight = 208.5) intensity reduces and splits
into a two-peak structure (see the arrows in the inset) due to the annealing process while the peak intensity change of oxygen seems to be negligible. Therefore the thermal energy applied by annealing may promote N migration toward either the interface or surface region of the layer [15]. For detailed information on the interface layer of annealed HfOyNz on Si, we measured electron energy loss spectroscopy (EELS). It was found that the N signal increases at the interface compared to the middle of the HfOyNz layer (not shown here). From both measurements we can conclude that the compositional change of N in the HfOyNz may result in a big layer thickness change, while at the same time a lower leakage current was maintained compared to where Si is incorporated in HfO2. The presence of impurities in dielectric layers not only contributes to the electrical properties but also to the crystalline structure. Residual carbon (C) was revealed to influence the leakage current in HfO2 films and can be moderated by ozone [16]. Another impurity, fluorine (F) affects the quality of the dielectric layer by passivating oxygen vacancies in high-k materials [17]. In spite of no F source in the pure chemical structure of precursors, the amount of fluorine via precursor flow in ALD process may be engaged in the high-k layer formation. It was reported that the F concentration in high-k layer was very sensitive to the deposition temperature [18]. However, in this experiment, the deposition temperature of high-k dielectric layers was fixed, therefore the amount of fluorine should be affected by post annealing process. Fig. 6 presents the annealing effect on impurities of the HfOyNz sample. The annealing process seems to reduce the amount of F impurities in the layer. Interfacial F decreases more than four times by annealing while the change of amount of C in dielectric layer is almost identical before and after annealing. 4. Conclusions
Fig. 5. Depth profile comparison between as-grown and annealed HfOyNz.
We investigated the effect of post deposition annealing on electrical, structural and compositional properties of Hf-based gate oxides on Si. The main leakage current mechanism of all as-grown HfO2, HfSixOy and HfOyNz and annealed samples was revealed as Poole–Frenkel conduction at low applied voltage and Fowler– Nordheim tunneling in high field. In spite of the 0.3 nm thickness reduction after annealing, the HfOyNz sample showed the lowest leakage current. Decomposition of the interfacial SiO2 by HfOyNz layer deposition and annealing was identified by TEM observations. The obtained bandgap of as-grown samples are 5.35 to 5.40 eV. Annealing decreases the bandgap of all samples by about 1 eV. The reduced signal peak of the HfOyNz layer after annealing may indicate N migration toward the surface and interface regions of the layer. For the impurity change in the HfOyNz layer, it was
360
J.-H. Kim et al. / Microelectronic Engineering 86 (2009) 357–360
found that PDA decreased more than four times the F content in high-k layer, while it had negligible effect on the amount of C.
[5] [6] [7] [8]
Acknowledgments [9]
The authors would like to thank SEMATECH (USA) and Engelmann in AMD for the sample preparation and the TEM measurement. The work described here has been funded in line with the technology funding for regional development (ERDF) of the European Union and by funds of the Free State of Saxony. It was also partially supported by the Korean Creative Research Initiatives Program (EAPap Actuator) of KOSEF/MEST. References [1] [2] [3] [4]
X. Zhao, D. Ceresoli, D. Vanderbilt, Phys. Rev. B 71 (2005) 085107. D. Ceresoli, D. Vanderbilt, Phys. Rev. B 74 (2006) 125108. T. Ono, K. Kato, H. Toyota, Y. Fukuda, Y. Jin, Jpn. J. Appl. Phys. 45 (2006) 7345. H. Wong, H. Iwai, Microelectron. Eng. 83 (2006) 1867.
[10] [11] [12] [13] [14] [15]
[16] [17] [18]
R. Puthenkovilankam, J.P. Chang, J. Appl. Phys. 96 (2004) 2701. G.-M. Rignanese, J. Phys.: Condens. Matter 17 (2005) R357. P.K. Park, S.-W. Kang, Appl. Phys. Lett. 89 (2006) 192905. H. Wong (Ed.), Proceedings of the 5th IEEE International Caracas Conference on Devices, Circuits and Systems, November 2004, 56. S. Sayan, S. Aravamudhan, B.W. Busch, W.H. Schulte, F. Cosandey, G.D. Wilk, T. Gustafsson, E. Garfunkel, J. Vac. Sci. Technol., A 20 (2002) 507. S. Kamiyama, T. Miura, Y. Nara, T. Arikado, Appl. Phys. Lett. 86 (2005) 222904. P.F. Lee, J.Y. Dai, K.H. Wong, H.L.W. Chan, C.L. Choy, J. Appl. Phys. 93 (2003) 3665. J. Robertson, Rep. Prog. Phys. 69 (2006) 327. M. Specht, M. Städele, S. Jakschik, U. Schröder, Appl. Phys. Lett. 84 (2004) 3076. S.J. Wang, P.C. Lim, A.C.H. Huan, C.L. Liu, Q. Li, C.K. Ong, Appl. Phys. Lett. 82 (2003) 2047. M. Queverdo-Lopez, M. El-Bouanani, S. Addepalli, J.L. Duggan, B.E. Gnade, R.M. Wallace, M.R. Visokay, M. Douglas, L. Colombo, Appl. Phys. Lett. 79 (2001) 4192. M. Cho, J.H. Kim, C.S. Hwang, H.-S. Ahn, S. Han, J.Y. Won, Appl. Phys. Lett. 90 (2007) 182907. T. Sasaki, Y. Akasaka, K. Miyagawa, T. Hoshi, Y. Watanabe, F. Ootsuka, M. Yasuhira, T. Arikado, Jpn. J. Appl. Phys. 44 (2005) 2252. J.-H. Kim, V. Ignatova, P. Kücher, J. Heitmann, L. Oberbeck, U. Schröder, Thin Solid Films 516 (2008) 8333.
Contents lists available at ScienceDirect
Microelectronic Engineering journal homepage: www.elsevier.com/locate/mee
Annealing effect on physical and electrical characteristics of thin HfO2, HfSixOy and HfOyNz films on Si Joo-Hyung Kim a,c,*, Velislava A. Ignatova a, Martin Weisheit b a
Fraunhofer Institute, Center of Nanoelectronic Technologies (CNT), Königsbrücker Str., 01099 Dresden, Germany Center for Complex Analysis, AMD Fab36 LLC & Co. KG, D-01109 Dresden, Germany c Dept. of Mechanical Engineering, Inha University, 253 Young-Hyun Dong, Nam Gu, Incheon, Republic of Korea b
a r t i c l e
i n f o
Article history: Received 13 June 2008 Received in revised form 30 October 2008 Accepted 7 November 2008 Available online 24 November 2008 Keywords: HfO2 Gate oxide High-k Annealing Leakage current Bandgap
a b s t r a c t We report the effect of annealing on electrical and physical characteristics of HfO2, HfSixOy and HfOyNz gate oxide films on Si. Having the largest thickness change of 0.3 nm after post deposition annealing (PDA), HfOyNz shows the lowest leakage current. It was found for both as-grown and annealed structures that Poole–Frenkel conduction is dominant at low field while Fowler–Nordheim tunneling in high field. Spectroscopic ellipsometry measurement revealed that the PDA process decreases the bandgap of the dielectric layers. We found that a decreasing of peak intensity in the middle HfOyNz layer as measured by Tof-SIMS may suggest the movement of N toward the interface region between the HfOyNz layer and the Si substrate during the annealing process. Crown Copyright Ó 2008 Published by Elsevier B.V. All rights reserved.
1. Introduction Aggressive downscaling of CMOS circuits produces high leakage current due to direct tunneling of electrons through the gate SiO2 layer (thickness <1.5 nm). According to the International Technology Roadmap for Semiconductors (ITRS), to keep the equivalent oxide thickness with low leakage current in the gate oxide, alternative high dielectric constant (high-k) oxides such as ZrO2, HfO2 and Ta2O5 have been considered as possible candidates for further downscaling [1–4]. As the best candidate for gate oxide application, HfO2 has a high dielectric constant (ranging from 20 to 49 for the monoclinic, cubic and tetragonal phase [5–7], a large band gap with sufficient band offset (>1.5 eV) and is thermally stable in contact with silicon substrates [8,9]. The main drawback of HfO2 is the low crystallization temperature (T < 400 °C) which can promote crystal boundaries within the layers to act as leakage paths and impurity getters during processing. Therefore it is an important issue to increase the kinetic stability while keeping the advantages of HfO2 for its adoption into semiconductor manufacturing. The first method to increase the crystallization temperature of HfO2 is to form a silicate of HfO2, * Corresponding author. Address: Dept. of Mechanical Engineering, Inha University, 253 Young-Hyun Dong, Nam Gu, Incheon, Republic of Korea. Tel.: +82 32 874 7325; fax: +82 32 832 7325. E-mail addresses: [email protected], [email protected] (J.-H. Kim).
but then the dielectric constant decreases [10]. The other option is to form an aluminate layer of HfO2 which has a rather high dielectric constant due to Al2O3 itself being a high-k material [11]. Up to date, however, few publications show the annealing effect on the electrical performance of the dielectric layer and especially the elemental depth profile in the layer. This report presents the effect of annealing on the electrical and physical characteristics of Hf-based gated oxides. 2. Experimental procedure HfO2, HfSixOy and HfOyNz films with thicknesses around 4 nm were deposited on HF cleaned 1200 n-type Si substrates by atomic layer deposition (ALD). Tetrakis(ethylmethylamino)-hafnium (TEMA-Hf) and Tetrakis(ethylmethylamino)-silicon (TEMA-Si) plus ozone were used as precursors. Ion implantation, reported to be successfully integrated into a traditional CMOS process of N2+ in ALD was introduced to form the HfOyNz layer. After deposition, in order to investigate the annealing effect on Hf-based gate oxides, one group of samples was subsequently annealed at 1000 °C, 10 s in N2 ambient. In order to facilitate electrical characterization, top electrodes made up of 400 nm thick Al dots with different diameters – 350, 600 and 1300 lm – were deposited by sputtering on both the as-grown and annealed samples. The electrical properties were characterized by a semiconductor parameter analyser (HP 4156 A). Capacitance–voltage measurement was performed
0167-9317/$ - see front matter Crown Copyright Ó 2008 Published by Elsevier B.V. All rights reserved. doi:10.1016/j.mee.2008.11.012
358
J.-H. Kim et al. / Microelectronic Engineering 86 (2009) 357–360
by a precision LCR meter (HP 4284A) with parallel mode (Cp) at 1 kHz. Structural investigation was done by Transmission Electron Microscopy (TEM, FEI TECNAI F30). To obtain detailed information about the annealing effect on the dielectric layer as well as the interfaces between layers, we used time-of-flight secondary ion mass spectrometry (ToF-SIMS, ION-TOF-V) with a 1 keV Cs+ primary ion beam at incident angle of 45° for elemental depth distribution and the interface composition. 3. Results and discussion The leakage current plot as a function of applied voltage of as-grown samples is shown in Fig. 1. The leakage current was obtained under electron injection from the Si substrate. For all asgrown samples the leakage current level is similar up to an applied voltage of 2.5 V while it becomes clearly different as the applied field is higher than 2.5 V. The highest leakage current was observed for the HfO2 sample in the whole range of applied fields. The HfSixOy and HfOyNz samples show a lower leakage current, which indicates Si and N incorporation into HfO2 has a role to decrease leakage current due to their high conduction band (CB) offset of 3.2 and 2.4 eV for SiO2 and Si3N4, respectively. Robertson [12] reports that the CB offset of HfSiO4 is 1.8 eV, which is between that of HfO2 (1.4 eV) and SiO2. The inset in Fig. 1 shows the two regions of different leakage current mechanisms of the HfOyNz sample: Trap assisted tunneling (TAT) and Poole–Frenkel (P–F) conduction in low electric field were observed [13] while Fowler–Nordheim (F–N) tunneling was dominant at high field. From the P–F plots in ln (J/E) vs. E1/2, the trap depth /PF of as-grown HfO2, HfSixOy and HfOyNz samples was 0.57 eV, 0.61 eV and 0.59 eV, respectively. It known that Si or N incorporation into HfO2 increases the trap depth of dielectric layers [5]. Fig. 2 presents the annealing effect on leakage current of Hfbased oxide samples. The leakage current of all annealed samples at 1 V is more than first order of magnitude higher than that of the as-grown samples. The leakage current mechanism is similar to the as-grown samples in Fig. 1, while the transition point from P–F to F–N occurs at higher field. Since the leakage current is related to the physical thickness and the band gap (BG) of a layer, we measured the thickness change and bandgap of as-grown and annealed samples by variable angle spectroscopic ellipsometry (Woollan VASE M-2000, spectral range 200–1000 nm). The measured data is summarized in Table 1. It is noteworthy that the biggest annealing effect on thickness
Fig. 1. Leakage current plot of as-grown HfO2 based metal–insulator–semiconductor (MIS) structures. The inset presents the P–F plot of the HfOyNz sample.
Fig. 2. Annealing effect on leakage current of HfO2 based MIS capacitors. The inset shows the region of Poole–Frenkel conduction (<1.0 V) and Fowler–Nordheim tunneling in high field range (>1.1 V). The crossover point from P–F to F–N is lower than for as-grown samples.
Table 1 Annealing effect on physical thickness of Hf-based gate oxides measured by spectroscopic ellipsometry.
As-grown Annealed
HfO2 (nm)
HfSixOy (nm)
HfOyNz (nm)
4.35 ± 0.02 4.25 ± 0.03
4.36 ± 0.02 4.17 ± 0.03
4.45 ± 0.02 4.15 ± 0.03
change was found in the HfOyNz sample. The dielectric functions of the films, which show a steep rise of absorption in the UV part of the spectrum, were parameterized by a Tauc-Lorentz-function. The bandgaps obtained from a fit of this parameterization to the ellipsometric data are 5.35, 5.40 and 5.38 eV for the as-grown HfO2, HfSixOy and HfOyNz, respectively. Regarding the annealing effect on the bandgap, shown in Fig. 3, it is clearly revealed that the annealing process promotes the decrease of the bandgap for all samples in this study. Wang et al reported that the formation of Si-rich Hf-silicate stems from the linkage between SiO2 and Hf precursor during the first HfO2 monolayer formation [14]. Since Hf-silicate is thermally stable in contact with Si and HfO2, silicate formation may occur during the deposition of HfO2 and HfOyNz. Additional annealing would promote silicate formation in the interface layer. To elucidate this point further, we characterized
Fig. 3. Bandgap changes of as-grown and annealed HfO2, HfSixOy and HfOyNz samples using variable angle spectroscopic ellipsometry (VASE).
J.-H. Kim et al. / Microelectronic Engineering 86 (2009) 357–360
359
Fig. 6. Annealing effect on carbon and fluorine contents of HfOyNz/SiO2/Si.
Fig. 4. Cross-sectional TEM image of the annealed HfOyNz/Si sample. The inset is a detail image of the interface between HfOyNz and Si.
the cross-section of the annealed HfOyNz sample by TEM. Fig. 4 shows that there is no amorphous layer formed at the interface, while the amorphous SiO2 layer before the dielectric layer deposition was around 1 nm. This indicates that the interfacial SiO2 layer seems to be decomposed during the dielectric layer deposition and annealing process [12]. The equivalent oxide thickness (EOT = thigh-kkSiO2/khigh-k, where thigh-k is the physical thickness of high-k layer, kSiO2 is the dielectric constant of SiO2, and khigh-k is the dielectric constant of high-k layer from capacitance–voltage measurement) was calculated from C–V measurements. The EOT values of as-grown Hf-based samples were 0.4–0.7 nm while those of the annealed samples were slight larger at around 0.5–0.6 nm. The dielectric constants were 31, 30 and 28 for the annealed HfO2, HfSixOy and HfOyNz, respectively, which are slightly higher than those of as-grown samples except HfSixOy. To get more detailed information about the chemical composition, we measured the elemental depth profile of the as-grown and annealed HfOyNz samples using secondary ion mass spectroscopy (SIMS). Fig. 5 shows the difference in the depth profiles of HfOyNz in as-grown and annealed samples. It was observed that the HfON secondary ion (atomic weight = 208.5) intensity reduces and splits
into a two-peak structure (see the arrows in the inset) due to the annealing process while the peak intensity change of oxygen seems to be negligible. Therefore the thermal energy applied by annealing may promote N migration toward either the interface or surface region of the layer [15]. For detailed information on the interface layer of annealed HfOyNz on Si, we measured electron energy loss spectroscopy (EELS). It was found that the N signal increases at the interface compared to the middle of the HfOyNz layer (not shown here). From both measurements we can conclude that the compositional change of N in the HfOyNz may result in a big layer thickness change, while at the same time a lower leakage current was maintained compared to where Si is incorporated in HfO2. The presence of impurities in dielectric layers not only contributes to the electrical properties but also to the crystalline structure. Residual carbon (C) was revealed to influence the leakage current in HfO2 films and can be moderated by ozone [16]. Another impurity, fluorine (F) affects the quality of the dielectric layer by passivating oxygen vacancies in high-k materials [17]. In spite of no F source in the pure chemical structure of precursors, the amount of fluorine via precursor flow in ALD process may be engaged in the high-k layer formation. It was reported that the F concentration in high-k layer was very sensitive to the deposition temperature [18]. However, in this experiment, the deposition temperature of high-k dielectric layers was fixed, therefore the amount of fluorine should be affected by post annealing process. Fig. 6 presents the annealing effect on impurities of the HfOyNz sample. The annealing process seems to reduce the amount of F impurities in the layer. Interfacial F decreases more than four times by annealing while the change of amount of C in dielectric layer is almost identical before and after annealing. 4. Conclusions
Fig. 5. Depth profile comparison between as-grown and annealed HfOyNz.
We investigated the effect of post deposition annealing on electrical, structural and compositional properties of Hf-based gate oxides on Si. The main leakage current mechanism of all as-grown HfO2, HfSixOy and HfOyNz and annealed samples was revealed as Poole–Frenkel conduction at low applied voltage and Fowler– Nordheim tunneling in high field. In spite of the 0.3 nm thickness reduction after annealing, the HfOyNz sample showed the lowest leakage current. Decomposition of the interfacial SiO2 by HfOyNz layer deposition and annealing was identified by TEM observations. The obtained bandgap of as-grown samples are 5.35 to 5.40 eV. Annealing decreases the bandgap of all samples by about 1 eV. The reduced signal peak of the HfOyNz layer after annealing may indicate N migration toward the surface and interface regions of the layer. For the impurity change in the HfOyNz layer, it was
360
J.-H. Kim et al. / Microelectronic Engineering 86 (2009) 357–360
found that PDA decreased more than four times the F content in high-k layer, while it had negligible effect on the amount of C.
[5] [6] [7] [8]
Acknowledgments [9]
The authors would like to thank SEMATECH (USA) and Engelmann in AMD for the sample preparation and the TEM measurement. The work described here has been funded in line with the technology funding for regional development (ERDF) of the European Union and by funds of the Free State of Saxony. It was also partially supported by the Korean Creative Research Initiatives Program (EAPap Actuator) of KOSEF/MEST. References [1] [2] [3] [4]
X. Zhao, D. Ceresoli, D. Vanderbilt, Phys. Rev. B 71 (2005) 085107. D. Ceresoli, D. Vanderbilt, Phys. Rev. B 74 (2006) 125108. T. Ono, K. Kato, H. Toyota, Y. Fukuda, Y. Jin, Jpn. J. Appl. Phys. 45 (2006) 7345. H. Wong, H. Iwai, Microelectron. Eng. 83 (2006) 1867.
[10] [11] [12] [13] [14] [15]
[16] [17] [18]
R. Puthenkovilankam, J.P. Chang, J. Appl. Phys. 96 (2004) 2701. G.-M. Rignanese, J. Phys.: Condens. Matter 17 (2005) R357. P.K. Park, S.-W. Kang, Appl. Phys. Lett. 89 (2006) 192905. H. Wong (Ed.), Proceedings of the 5th IEEE International Caracas Conference on Devices, Circuits and Systems, November 2004, 56. S. Sayan, S. Aravamudhan, B.W. Busch, W.H. Schulte, F. Cosandey, G.D. Wilk, T. Gustafsson, E. Garfunkel, J. Vac. Sci. Technol., A 20 (2002) 507. S. Kamiyama, T. Miura, Y. Nara, T. Arikado, Appl. Phys. Lett. 86 (2005) 222904. P.F. Lee, J.Y. Dai, K.H. Wong, H.L.W. Chan, C.L. Choy, J. Appl. Phys. 93 (2003) 3665. J. Robertson, Rep. Prog. Phys. 69 (2006) 327. M. Specht, M. Städele, S. Jakschik, U. Schröder, Appl. Phys. Lett. 84 (2004) 3076. S.J. Wang, P.C. Lim, A.C.H. Huan, C.L. Liu, Q. Li, C.K. Ong, Appl. Phys. Lett. 82 (2003) 2047. M. Queverdo-Lopez, M. El-Bouanani, S. Addepalli, J.L. Duggan, B.E. Gnade, R.M. Wallace, M.R. Visokay, M. Douglas, L. Colombo, Appl. Phys. Lett. 79 (2001) 4192. M. Cho, J.H. Kim, C.S. Hwang, H.-S. Ahn, S. Han, J.Y. Won, Appl. Phys. Lett. 90 (2007) 182907. T. Sasaki, Y. Akasaka, K. Miyagawa, T. Hoshi, Y. Watanabe, F. Ootsuka, M. Yasuhira, T. Arikado, Jpn. J. Appl. Phys. 44 (2005) 2252. J.-H. Kim, V. Ignatova, P. Kücher, J. Heitmann, L. Oberbeck, U. Schröder, Thin Solid Films 516 (2008) 8333.