Ozone-Based Metal Oxide Atomic Layer Deposition

Electrochemical and Solid-State Letters, 13 共6兲 H176-H178 共2010兲

H176

1099-0062/2010/13共6兲/H176/3/$28.00 © The Electrochemical Society

Ozone-Based Metal Oxide Atomic Layer Deposition: Impact of N2 ÕO2 Supply Ratio in Ozone Generation Annelies Delabie,a,z Matty Caymax,a,* Sven Gielis,a Jan Willem Maes,b Laura Nyns,a,** Mihaela Popovici,a Johan Swerts,a Hilde Tielens,a Jozef Peeters,c and Sven Van Elshochta,* a

Imec, B-3001 Leuven, Belgium ASM Belgium, B3001 Leuven, Belgium Department of Chemistry, University of Leuven, B-3001 Leuven, Belgium

b c

The O2 /N2 flow ratio during O3 generation by dielectric barrier discharge has a large impact on the atomic layer deposition 共ALD兲 of metal oxides in a hot wall ALD reactor. For HfO2 ALD using HfCl4 as a metal precursor, a higher growth per cycle and a broader ALD temperature window are obtained when N2 is added to the O2 supply of the O3 generation. A positive impact of N2 in the O3 generation is also observed for ZrO2 and La2O3 ALD. A negative impact is observed for Al2O3 ALD: The Al2O3 thickness is reduced for those conditions for O3 where HfO2 ALD is enhanced. © 2010 The Electrochemical Society. 关DOI: 10.1149/1.3355207兴 All rights reserved. Manuscript submitted January 15, 2010; revised manuscript received February 12, 2010. Published March 18, 2010.

Atomic layer deposition 共ALD兲 is a thin-film deposition technique that uses self-limiting chemisorption reactions of gas-phase precursors. Many metal oxides have been deposited by ALD1 using H2O or O3 as oxidant precursors. The use of O3 has several advantages. O3 is a strong oxidant. Therefore, some metal precursors react with O3 and not with H2O, e.g., ␤-diketonato compounds.1 At low temperature, O3 is easier to purge from the reactor than H2O, which slowly desorbs from the reactor walls, as such introducing a contribution of gas-phase reactions to the ALD. Due to the high reactivity and easy purging, O3 allows ALD of metal oxides at low temperatures.2 Furthermore, O3 can be preferred for deposition of or on hygroscopic materials, e.g., ALD of rare-earth oxides3 or high-k dielectric oxide ALD on GeO2 passivation layers in complementary metal oxide semiconductor devices with high mobility channel materials.4 Finally, film nucleation on less reactive surfaces might proceed easier with O3 than with H2O because O3 can oxidize the starting surface.5,6 O3 is frequently generated from pure O2, N2-doped O2, or dry air 共80% N2 /20% O2兲 by a dielectric-barrier discharge.7-10 The first step toward O3 formation is electron impact dissociation of O2 共O2 + e− → 2 O + e−兲.10 O3 is then formed in a three-body reaction involving O and O2 共O + O2 + M → O3 + M, with M as a third collision partner兲. In the presence of N2, discharge products such as NO, N2O, NO2, NO3, N2O5 have also been identified.8-10 For example, with 80% N2, the stable oxides N2O and N2O5 were detected in concentrations 2 orders of magnitude lower than for O3.8 In this work, we investigate the effect of the N2 /O2 flow ratio during O3 generation on the ALD of oxides. Although there are articles on HfO2 layers deposited by HfCl4 /O3 ALD,4,11-13 the process characteristics are scarcely documented. We have investigated HfCl4 /O3 ALD in hot wall cross-flow ALD reactors and found a major impact of the O2 /N2 flow ratio during the O3 generation on the growth per cycle and ALD temperature window. Adding N2 during O3 generation also affects other processes such as ZrO2, La2O3, and Al2O3 ALD. The latter two processes are documented in literature,2,3,14-17 although the effect of the N2 /O2 ratio during the O3 generation has not been described. Si共100兲 substrates 共200 or 300 mm兲 were cleaned in an O3 /H2O solution, resulting in ⬃1.0 nm SiO2 layers.18 ALD was performed in hot wall cross-flow Pulsar 2000 and 3000 reactors.d The pressure

* Electrochemical Society Active Member. ** Electrochemical Society Student Member. z

E-mail: [email protected]

d

Pulsar and Polygon are trademarks of ASM International, The Netherlands.

in the reactor was ⬃1 Torr, and the temperature was varied between 200 and 350°C. For Pulsar 2000, O3 was generated from pure O2 in a TMEIC OP-250H O3 generator, providing 350 g/m3 O3. For Pulsar 3000, O3 was generated from an O2 /N2 mixture in an MKS ASTeX 8403 system, integrated in an IN-USA ozone delivery system 共model ODS-LF-1兲. The O2 /N2 flow ratio was varied from 0 to 20% N2 at a fixed O3 concentration of 233 g/m3. Also, 350 g/m3 O3 was used for comparison. The O3 flow in the ALD reactors was typically 500–600 sccm, but to check for saturation, flows up to 2 slm were considered. The O3 pulse time was varied between 250 ms and 15 s. For HfO2, ZrO2, Al2O3, and La2O3 ALD, HfCl4, ZrCl4, Al共CH3兲3 关trimethylaluminium 共TMA兲兴, and La共thd兲3 共thd = 2,2,6,6-tetramethyl-3,5-heptanedione兲, respectively, were used as metal precursors. TMA was delivered from a container at 18°C, while the other 共solid兲 precursors were delivered from containers heated to 165–190°C 共HfCl4 and ZrCl4兲 and 200°C 关La共thd兲3兴. Thicknesses 共typically 5–15 nm兲 were measured by spectroscopic ellipsometry. Time of flight secondary-ion mass spectroscopy 共TOF-SIMS兲 depth profiles were obtained using a dual ion beam setup with a 500 eV Xe+ ion beam, followed by 15 keV Ga+. The Hf content was measured by Rutherford backscattering spectroscopy 共RBS兲 in a Charles Evans & Associates RBS400 end station with a 1 MeV He+ beam in a rotating random mode. The HfO2 growth per cycle and temperature window in a hot wall ALD reactor are affected by the O2 /N2 flow ratio used during O3 generation 共Fig. 1 and 2兲. When O3 is generated from pure O2, a narrow ALD temperature window 共200–250°C兲 was observed in a Pulsar 2000 reactor. High quality, uniform HfO2 layers were deposited using large O3 doses 共10 s, 2 slm O3 /O2兲. At 225°C, the growth per cycle was 0.12 nm/cycle, and the thickness nonuniformity was below a 3% standard deviation. The nonuniformity of the HfO2 layers over the 200 mm substrates increased with temperature. The HfO2 thickness decreased over the substrate in the direction of the gas flow in the reactor 共e.g., from 0.08 to 0.01 nm/cycle at 300°C兲. Reactor temperatures above 250°C resulted in a too large nonuniformity 共⬎10% standard deviation兲, which was not improved by increasing the O3 dose 共e.g., Fig. 2, effect of O3 pulse time at 300°C兲 or the HfCl4 dose 共by increasing HfCl4 source temperature by 5 or 10°C兲. Reactor temperatures below 200°C were not considered as the reactor should be at a higher temperature than the HfCl4 source. In a Pulsar 3000 reactor, nonuniform HfO2 layers were obtained even at low temperatures when O3 was generated from pure O2 共Fig. 1兲, as discussed further below. To further adjust the oxidant power, O3 was generated in N2 /O2 mixtures, as such providing not only O3 but also, according to the well-established gas-phase chemistry,19 NO2 and NO3 species in the

Downloaded 19 Mar 2010 to 146.103.254.11. Redistribution subject to ECS license or copyright; see http://www.ecsdl.org/terms_use.jsp


Thickness per cycle (nm)

0.3

0%N2 PULSAR2000 0%N2 PULSAR3000 20% N2 PULSAR3000

0.25 0.2

0.15 0.1

0.05 0 175

200

225

250

275

300

325 350

Deposition Temperature (°C)

Figure 1. 共Color online兲 HfO2 thickness per cycle as a function of deposition temperature for HfO2 depositions in Pulsar 2000 共100 cycles兲 or Pulsar 3000 reactors 共50 cycles兲. O3 was generated from pure O2 共0% N2兲 or from a N2 /O2 mixture 共20% N2兲. A constant O3 pulse time of 10 s was used. The HfO2 thickness was measured by ellipsometry on 49 points on the 200 共for Pulsar 2000兲 or 300 mm 共for Pulsar 3000兲 Si wafers. The mean thickness per cycle is represented on the chart. Error bars indicate the maximum and minimum thicknesses per cycle on the wafer.

HfO2 thickness/cycle (nm)

ALD reactor. With 20% N2 /O2, a broad ALD temperature window 共225–300°C兲 with high growth per cycle was observed in a Pulsar 3000 reactor. Uniform HfO2 films were obtained using the 共standard兲 500 sccm O3 flow 共Fig. 1兲. The best uniformity was observed at 300°C 共⬍1% standard deviation兲. The growth per cycle at 300°C was 0.15 nm/cycle, which is three times higher than that for HfCl4 /H2O ALD 共0.05–0.06 nm/cycle兲. Typical ALD saturation of the O3 reaction was observed 共Fig. 2a兲. At least 20% N2 was needed for uniform HfO2 layers 共Fig. 2b兲. TOF-SIMS depth profiles indicated low impurity content in the HfO2 layers deposited at 225°C with pure O3 and at 300°C with O3 produced in 20% N2 /O2. The Cl content was comparable to that in

0.16 0.14

20%

0.12 0.10 0.08 0.06 0.04

4% 0%

0.02 0

(a)

10 s O3

0.00 5

10

15 0%

O3 pulse length (s)

10%

(b) N2/O2 %

20%

Figure 2. 共Color online兲共a兲 HfO2 thickness per cycle as a function of the O3 pulse time for different N2 /O2 flow ratios used during O3 generation; 共b兲 HfO2 thickness per cycle as a function of the N2 /O2 flow ratio during O3 generation at a fixed O3 pulse time of 10 s. The HfO2 depositions 共50 cycles兲 were performed in a Pulsar 3000 reactor at 300°C. The HfO2 thickness was measured by ellipsometry on 49 points on a 300 mm Si wafers. The mean thickness per cycle is represented on the chart. Error bars indicate the maximum and minimum thicknesses per cycle on the wafer.

H177

HfO2 deposited by HfCl4 /H2O ALD at 300°C.20,e No significant amount of N was present: Both CN− and C− signals were at the detection limit. These HfO2 layers exhibited good dielectric properties on Si and Ge substrates.4,21 In contrast, the HfO2 layers deposited at 225°C with O3 produced in 20% N2 /O2 indicated high levels of Cl and N and poor dielectric properties.21 For O3 generation in pure O2, the nonuniformity of HfO2 at a high reactor temperature 共⬎250°C兲 is caused by the decomposition of O3 at the hot HfO2-coated reactor walls. This was demonstrated by oxidizing H-terminated Si wafers by O3 in the ALD reactor, where the reactor walls were coated by a HfO2 layer by ALD. After a 30 min oxidation at 300°C, the SiO2 thickness decreased in the direction of the gas flow from 1 to 0.7 nm, indicating a decreasing concentration of O3. The minimum SiO2 thickness 共0.7 nm兲 was equal to that obtained by oxidation in O2, indicating that no significant amount of O3 reached the end of the reactor. The same SiO2 thickness gradient was observed when the reactor walls were coated by ZrO2. In contrast, uniform oxidation to 1.2 nm SiO2 was observed when the reactor walls were coated by Al2O3. Thus, O3 decomposition occurs on the hot HfO2 and ZrO2 surfaces, resulting in a decreasing O3 concentration along the gas stream through the cross-flow reactor. Decomposition of O3 was previously proposed to account for long O3 pulse times necessary for ALD of MnO2, a well-known catalyst for the decomposition of O3,22 and to explain the O3 oxidation kinetics of Ge in a Pulsar 3000 reactor.21 The effect of O3 wall destruction depends on the reactor design: In Pulsar 3000, the HfCl4 and O3 gas lines are merged more upstream of the wafer than in Pulsar 2000. Therefore, in Pulsar 3000, O3 already starts to decompose on the HfO2 coating before the wafer, resulting in worse HfO2 uniformities. The SiO2 thickness gradient decreased at lower reactor temperatures, indicating a longer lifetime of O3, enabling HfCl4 /O3 ALD at a low temperature in Pulsar 2000. Keeping the reactor walls at a lower temperature than the wafer helps to minimize O3 decomposition. The HfO2 共ZrO2兲 coating in the reactor can be partially passivated by extended O3 exposures 共several hours兲, but it is not practical to include such long treatments in each reaction cycle. We have not identified the reaction mechanism of O3-based HfO2 ALD. The addition of N2 can result in a longer lifetime of O3 in the ALD reactor through passivation of the reactor walls by NO2 or NO3. NO2 and NO3 likely contribute in the ALD chemisorption reactions as they are strong oxidants. Our HfO2 ALD characteristics show remarkable similarities with those reported for N2O/O2 plasma-based ALD.23-25 The addition of small amounts of N2O in O2 plasma enhanced the growth per cycle of HfO2 and ZrO2 ALD with tetrakis ethylmethylamino hafnium and tetrakis ethylmethylamino zirconium, resulting in similarly high growth per cycle values 共0.14 nm/cycle for HfO2兲. Intermediate N-related surface species were proposed to play a role in enhancing the growth per cycle. The addition of N2 in the O3 generation is also beneficial for the La共thd兲3 /O3 ALD of La2O3. It results in a higher throughput process: Uniform films were obtained at shorter O3 pulse times when N2 /O2 was used in the O3 generation.f In contrast, the addition of N2 in the O3 generation has a negative impact on TMA/O3 ALD of Al2O3. It results in the unconventional saturation behavior of the O3 pulse time 共Fig. 3a for TMA/O3 ALD at 300°C兲. Increasing the O3 pulse time first resulted in a thickness increase 共as expected兲, but then the thickness decreased for longer O3 pulse times. In contrast, a normal saturation behavior was observed with pure O3 共Fig. 3a兲. The TMA/O3 共N2 /O2兲 ALD resulted e

For HfCl4 /H2O ALD, the Cl content strongly depends on the H2O pulse time, which can vary from 300 to 500 ms in high throughput processes of up to 10 s for research applications.21 The Cl content of HfCl4 /O3 共N2 /O2兲 ALD was lower than that of HfCl4 /H2O ALD with H2O pulse times of 300–500 ms but higher than that of HfCl4 /H2O ALD with 10 s H2O pulses. f The effect of N2 in the generation of O3 for La2O3 ALD is the subject of further investigation and is therefore not included in this article.



0.10

In summary the N2 /O2 flow ratio during the O3 generation in a dielectric barrier discharge strongly affects the ALD process characteristics in a hot wall cross-flow ALD reactor. As it is a significant process parameter, documentation is required. Further studies are needed to investigate the reaction mechanism of O3-based ALD and the contributions of NO2 and NO3.

0.08 0.06 0.095 0.04 0.02

Acknowledgments

0.085 0.075 0

0.00 0

1

5

10

15

O3 pulse length (s)

0.12

3

0.08

2

0.04

1

0 0.00

(b)

0.5

2

Al2O3 thickness/cycle (nm)

(a)

20% N2/O2 4% N2/O2 1.2% N2/O2 pure O2

Hf-content (/nm )

Al2O3 thickness/cycle (nm)

H178

0.05

0.10

N2/O2

0.15

0 0.20

Figure 3. 共a兲 Al2O3 thickness per cycle as a function of the O3 pulse time for different N2 /O2 flow ratios used during O3 generation. The Al2O3 depositions 共100 cycles兲 were performed in a Pulsar 3000 reactor at 300°C. The Al2O3 thickness was measured by ellipsometry on 49 points on a 300 mm Si wafers. The mean thickness per cycle is represented on the chart. Error bars indicate the maximum and the minimum thickness per cycle on the wafer. 共b兲 Al2O3 thickness per cycle as a function of the N2 /O2 flow ratio in the O3 generation at a fixed O3 pulse time of 10 s; Hf content obtained by a single HfCl4 chemisorption reaction on 10 nm Al2O3 layers grown with different N2 /O2 flow ratios in the generation of O3 at a fixed O3 pulse time of 10 s, measured by RBS.

in N impurities in the films as detected by TOF-SIMS, whereas the C content was close to the detection limit. Thus, whereas using a N2 /O2 mixture during the O3 generation enhances surface reactions in HfO2 ALD, it reduced the Al2O3 thickness per cycle 共Fig. 2a vs Fig. 3a兲. The difference in the reactivities of HfCl4 and TMA was further investigated. The HfCl4 chemisorption reaction strongly depends on the type and number of surface sites.26,27 To compare the reactivity of Al2O3 surfaces obtained with different N2 /O2 ratios in the O3 generation, we compare the Hf content obtained after a single HfCl4 chemisorption reaction on those Al2O3 surfaces. The Hf content increased when the N2 /O2 flow ratio increased 共Fig. 3b兲, indicating that surface species that enhance HfCl4 chemisorption might block TMA chemisorption.

We kindly acknowledge TMEIC 共Toshiba-Mitsubishi Electric Industrial Systems Corporation兲 for providing an O3 generator, and Alexis Franquet 共Imec兲 and Bert Brijs 共Imec兲 for the TOF-SIMS and RBS measurements, respectively. Imec assisted in meeting the publication costs of this article.

References 1. R. L. Puurunen, J. Appl. Phys., 97, 121301 共2005兲. 2. S. K. Kim, S. W. Lee, C. S. Hwang, Y. S. Min, J. Y. Won, and J. Jeong, J. Electrochem. Soc., 153, F69 共2006兲. 3. N. Nieminen, M. Putkonen, and L. Niinisto, Appl. Surf. Sci., 174, 155 共2001兲. 4. A. Delabie, A. Alian, F. Bellenger, M. Caymax, T. Conard, A. Franquet, S. Sioncke, S. Van Elshocht, M. M. Heyns, and M. Meuris, J. Electrochem. Soc., 156, G163 共2009兲. 5. P. D. Kirsch, M. A. Quevedo-Lopez, H.-J. Li, Y. Senzaki, J. J. Peterson, S. C. Song, S. A. Krishnan, N. Moumen, J. Barnett, G. Bersuker, et al., J. Appl. Phys., 99, 023508 共2006兲. 6. B. Lee, S.-Y. Park, H.-C. Kim, K. Cho, E. M. Vogel, M. J. Kim, R. M. Wallace, and J. Kima, Appl. Phys. Lett., 92, 203102 共2008兲. 7. J. Kitayama and M. Kuzumoto, J. Phys. D: Appl. Phys., 32, 3032 共1999兲. 8. B. Eliasson and U. Kogelschatz, IEEE Trans. Plasma Sci., 19, 309 共1991兲. 9. Y. Nomoto, T. Ohkubo, S. Kanazawa, and T. Adachi, IEEE Trans. Ind. Appl., 31, 1458 共1995兲. 10. U. Kogelschatz, B. Eliasson, and W. Egli, Pure Appl. Chem., 71, 1819 共1999兲. 11. S. Baldovino, S. Spiga, G. Scarel, and M. Fanciulli, Appl. Phys. Lett., 91, 172905 共2007兲. 12. H. B. Park, M. Cho, J. Park, S. W. Lee, C. S. Hwang, J.-P. Kim, J.-H. Lee, N.-I. Lee, H.-K. Kang, J.-C. Lee, et al., J. Appl. Phys., 94, 3641 共2003兲. 13. S. Spiga, C. Wiemer, G. Tallarida, G. Scarel, S. Ferrari, G. Seguini, and M. Fanciulli, Appl. Phys. Lett., 87, 112904 共2005兲. 14. D. N. Goldstein, J. A. McCormick, and S. M. George, J. Phys. Chem. C, 112, 19530 共2008兲. 15. S. D. Elliott, G. Scarel, C. Wiemer, M. Fanciulli, and G. Pavia, Chem. Mater., 18, 3764 共2006兲. 16. J. B. Kim, D. R. Kwon, K. Chakrabarti, C. Lee, K. Y. Oh, and J. H. Lee, J. Appl. Phys., 92, 6739 共2002兲. 17. S. K. Kim and C. S. Hwang, J. Appl. Phys., 96, 2323 共2004兲. 18. M. Meuris, P. W. Mertens, A. Opdebeeck, H. F. Schmidt, M. Depas, G. Vereecke, M. M. Heyns, and A. Philipossian, Solid State Technol., 38, 109 共1995兲. 19. R. Atkinson, D. L. Baulch, R. A. Cox, R. F. Hampson, J. A. Kerr, M. J. Rossi, and J. Troe, J. Phys. Chem. Ref. Data, 26, 521 共1997兲. 20. A. Delabie, M. Caymax, B. Brijs, D. P. Brunco, T. Conard, E. Sleeckx, S. Van Elshocht, L.-Å. Ragnarsson, S. De Gendt, and M. M. Heyns, J. Electrochem. Soc., 153, F180 共2006兲. 21. A. Delabie, A. Alian, F. Bellenger, G. Brammertz, D. P. Brunco, M. Caymax, T. Conard, A. Franquet, M. Houssa, S. Sioncke, et al., ECS Trans, 16共10兲, 671 共2008兲. 22. O. Nilsen, H. Fjellvag, and A. Kjekshus, Thin Solid Films, 444, 44 共2003兲. 23. S.-J. Won, J.-Y. Kim, G.-J. Choi, J. Heo, C. S. Hwang, and H. J. Kim, Chem. Mater., 21, 4374 共2009兲. 24. G.-J. Choi, S. K. Kim, S.-J. Won, H. J. Kim, and C. S. Hwang, J. Electrochem. Soc., 156, G138 共2009兲. 25. S. Kim, J. Kim, J. Choi, H. Kang, H. Jeon, and C. Bae, J. Vac. Sci. Technol. B, 24, 1088 共2006兲. 26. M. L. Green, M.-Y. Ho, B. Busch, G. D. Wilk, T. Sorsch, T. Conard, B. Brijs, W. Vandervorst, P. I. Räisänen, D. Muller, et al., J. Appl. Phys., 92, 7168 共2002兲. 27. L. Nyns, A. Delabie, M. Caymax, M. M. Heyns, S. Van Elshocht, C. Vinckier, and S. De Gendt, J. Electrochem. Soc., 155, G269 共2008兲.
