Langmuir 2005, 21, 7321-7325
7321
In Situ Quadrupole Mass Spectrometry Study of Atomic-Layer Deposition of ZrO2 Using Cp2Zr(CH3)2 and Water Jaakko Niinisto¨,*,† Antti Rahtu,‡,§ Matti Putkonen,† Mikko Ritala,‡ Markku Leskela¨,‡ and Lauri Niinisto¨† Laboratory of Inorganic and Analytical Chemistry, Helsinki University of Technology, P.O. Box 6100, FIN-02015 Espoo, Finland, and Laboratory of Inorganic Chemistry, University of Helsinki, P.O. Box 55, FIN-00014 University of Helsinki, Finland Received January 10, 2005. In Final Form: May 26, 2005 Reactions during the atomic layer deposition (ALD) process of ZrO2 from Cp2Zr(CH3)2 and deuterated water as precursors were studied with a quadrupole mass spectrometer (QMS) at 210-440 °C. The detected reaction byproducts were CpD (m/z ) 67) and CH3D (m/z ) 17). Almost all (90%) of the CH3 ligands were released during the Cp2Zr(CH3)2 precursor pulse because of exchange reactions with the OD-terminated surface, and the rest, during the D2O pulse. About 40% of the CpD was released during the metal precursor pulse, and 60%, during the D2O pulse. ALD-type self-limiting growth was confirmed from 210 to 400 °C. However, below 300 °C the growth rate was low. Precursor decomposition affected the film growth mechanism at temperatures exceeding 400 °C.
Introduction Because of its attractive properties such as high permittivity (k ) 14-25),1 large band gap (5-7.8 eV),1,2 and thermodynamic stability on Si,3 zirconium oxide has gained considerable interest in the microelectronics industry as an alternative high-k oxide to replace the SiO2 gate dielectric in complementary metal oxide semiconductor devices.1 To produce the required highly conformal and uniform ultrathin films with accurate thickness control over a large surface area, atomic layer deposition (ALD) is considered to be the method of choice for high-k oxide deposition.4 ALD is based on sequential saturative surface reactions of the alternately applied precursors separated by purging gas pulses, thereby resulting in selflimiting film growth with accurate thickness control as well as excellent conformity.5-7 ZrO2 thin films have previously been grown by ALD using ZrCl4 most commonly as the metal precursor and water as the oxygen source.8-11 Unfortunately, this thoroughly studied process has some distinct drawbacks,
such as chlorine contamination and generation of corrosive HCl as a reaction byproduct. The problems with the ZrCl4/ H2O ALD process have motivated the search for alternative processes. Recently, it has been shown that organometallic-type precursors, viz., cyclopentadienyl (Cp, -C5H5)-based compounds, can be applied in the ALD of oxides (e.g., Cp2Zr(CH3)2 together with water yielded stoichiometric ZrO2 films with extremely low impurity contents (below 0.1 at. % for C and H) and promising electrical properties.12,13) Similar results have also been obtained with ALD processing of HfO2 using the analogous Cp compound, namely, Cp2Hf(CH3)2.14 Mass spectrometric studies, performed under similar conditions as used in the actual film depositions, offer the possibility to study the surface reactions in ALD.15-17 The aim of the present study is to exploit Cp2Zr(CH3)2/H2O process monitored in situ by quadrupole mass spectrometry (QMS) in a flow-type ALD reactor and thereby evaluate the usefulness of the cyclopentadienyl-type precursors. Experimental Section
* To whom correspondence should be addressed. E-mail:
[email protected]. † Helsinki University of Technology. ‡ University of Helsinki. § Present address: ASM Microchemistry Ltd, Ha ¨ meentie 135 A, FIN-00560 Helsinki, Finland. (1) Houssa, M.; Heyns, M. M. In High-k Gate Dielectrics; Houssa, M., Ed.; Institute of Physics Publishing: Bristol, U.K., 2004; p 10. (2) Wilk, G. D.; Wallace, R. M.; Anthony, J. M. J. Appl. Phys. 2001, 89, 5243. (3) Hubbard, K. J.; Schlom, D. G. J. Mater. Res. 1996, 11, 2757. (4) Hand, A. Semicond. Int. 2003, 26(5), 46. (5) Niinisto¨, L.; Ritala, M.; Leskela¨, M. Mater. Sci. Eng. B 1996, 41, 23. (6) Ritala, M.; Leskela¨, M. In Handbook of Thin Film Materials; Nalwa, H. S., Ed.; Academic Press: San Diego, CA, 2001; Vol. 1, pp 103-159. (7) Niinisto¨, L.; Pa¨iva¨saari, J.; Niinisto¨, J.; Putkonen, M.; Nieminen, M. Phys. Status Solidi A 2004, 201, 1443. (8) Ritala, M.; Leskela¨, M. Appl. Surf. Sci. 1994, 75, 333. (9) Copel, M.; Gribelyuk, M.; Gusev, E. Appl. Phys. Lett. 2000, 76, 436. (10) Aarik, J.; Aidla, A.; Ma¨ndar, H.; Uustare, T.; Sammelselg, V. Thin Solid Films 2002, 408, 97.
Experiments were carried out with a specially modified18,19 commercial flow-type F-120 ALD reactor manufactured by ASM Microchemistry Ltd. The reaction chamber was loaded with glass substrates so that these formed narrow flow channels between each other. The total area of the glass substrates was about (11) Nohira, H.; Tsai, W.; Besling, W.; Young, E.; Petry, J.; Conard, T.; Vandervorst, W.; De Gendt, S.; Heyns, M.; Maes, J.; Tuominen, M. J. Non-Cryst. Solids 2002, 303, 83. (12) Putkonen, M.; Niinisto¨, J.; Kukli, K.; Sajavaara, T.; Karppinen, M.; Yamauchi, H.; Niinisto¨, L. Chem. Vap. Deposition 2003, 9, 207. (13) Niinisto¨, J.; Putkonen, M.; Niinisto¨, L.; Kukli, K.; Ritala, M.; Leskela¨, M. J. Appl. Phys. 2004, 95, 84. (14) Niinisto¨, J.; Putkonen, M.; Niinisto¨, L. The AVS Topical Conference on Atomic Layer Deposition (ALD 2004); extended abstract on CD-ROM. (15) Juppo, M.; Rahtu, A.; Ritala, M.; Leskela¨, M. Langmuir 2000, 16, 4034. (16) Rahtu, A.; Alaranta, T.; Ritala, M. Langmuir 2001, 17, 6506. (17) Rahtu, A.; Ritala, M. J. Mater. Chem. 2002, 12, 1484. (18) Rahtu, A.; Ritala, M. Chem. Vap. Deposition 2002, 8, 21. (19) Rahtu, A.; Ritala, M. Electrochem. Soc. Proc. 2000, 2000-13, 105.
10.1021/la0500732 CCC: $30.25 © 2005 American Chemical Society Published on Web 07/06/2005
7322
Langmuir, Vol. 21, No. 16, 2005
3500 cm2. The gas composition was measured with a Hiden HAL/ 3F 501 RC QMS using an electron multiplier detector, a mass range of 1-510 amu, and an ionization energy of 70 eV. The sampling and the pressure reduction were accomplished through a 200 µm orifice. The pressure in the ALD reactor was about 1 mbar, and that in the QMS chamber was below 10-6 mbar. Cp2Zr(CH3)2, synthesized by the method described by Samuel and Rausch,20 was held inside the reactor in an open boat at 70 °C, and the pulsing was accomplished with inert gas valving.21 Cp2Zr(CH3)2 is moisture-sensitive; therefore, the precursor boat was loaded in a glovebox and exposed to air only for a couple of seconds before inserting it into the reactor. D2O (Euriso-top, 99.9% D) was held outside the reactor in a glass bottle at room temperature, the flow rate was controlled by a needle valve, and the pulsing was accomplished by a solenoid valve. It has been our common practice to use D2O in the in situ studies on processes involving organometallic precursors16,18,19 to better distinguish the reaction byproducts from the species forming in the QMS ionizator. Weak background signals were also arising even when no exchange reactions should have taken place (i.e., when subsequent pulses of only one precursor were given). Therefore, the background was subtracted from the data, as described earlier.22 Argon (99.999%) was used as a purging and carrier gas. Reaction temperatures were 210-440 °C. The precursor pulse times were varied between 3 and 6 s. The purge time was 5-10 s after each precursor pulse.
Results and Discussion The fragmentation of Cp2Zr(CH3)2 in the electron ionizer was reported earlier by Codato et al.23 In the present study, the ion with the highest m/z value was detected at m/z ) 235, which corresponds to the loss of a methyl radical; further methyl loss yielded an even stronger intensity peak at m/z ) 220. This [Cp2Zr]+ radical cation has been reported as being the most abundant species in the electron ionization spectra of (C5H5)2Zr(CH3)223 and zirconium cyclopentadienyl carbonyls.24 In addition to the abovementioned earlier study,23 where only higher m/z value (>120) peaks were reported, peaks originating from ions [Zr]+, [C5H5]+, [C2H6]+, and [CH3]+ were also detected. The molecular peak of (C5H5)2Zr(CH3)2 (m/z ) 250) was not detected at all, however. Thermal decomposition of the Zr precursor was studied by omitting water and supplying only repeated (C5H5)2Zr(CH3)2 pulses, separated by purge periods. Prior to the experiment, the surface was saturated with the precursor; that is, the surface was covered with -Cp2 - xZr(CH3)2 - y. The zirconium-containing species of the precursor (e.g., m/z ) 235, 220, 155, or 90) and the ligand species (Cp+, m/z ) 65) were monitored with QMS as a function of substrate temperature. The amounts of these species stayed rather constant at 200-375 °C. At higher temperatures, the amounts of zirconium-containing species considerably decreased whereas that of Cp+ (m/z ) 65) remained constant. This decomposition behavior can be seen in Figure 1, where the ratios of intensities of the Cp ligand and the [CpZr]+ or [Zr]+ ion are plotted as a function of temperature. The decomposition onset temperature of the zirconium precursor is about 375 °C. A slightly higher decomposition temperature has been detected earlier in thin film deposition experiments where the growth rate of ZrO2 in the Cp2Zr(CH3)2/H2O ALD process increased rapidly when the deposition temperature exceeded 400 °C, also resulting in a thickness profile that indicates (20) Samuel, E.; Rausch, M. D. J. Am. Chem. Soc. 1973, 95, 6263. (21) Suntola, T. Thin Solid Films 1992, 216, 84. (22) Rahtu, A.; Kukli, K.; Ritala, M. Chem. Mater. 2001, 13, 817. (23) Codato, S.; Carta, G.; Rossetto, G.; Zanella, P.; Gioacchini, A. M.; Traldi, P. Rapid Commun. Mass Spectrom. 1998, 12, 1981. (24) Thomas, J. L.; Brown, K. T. J. Organomet. Chem. 1976, 111, 297.
Niinisto¨ et al.
Figure 1. Ratio between the intensities of the signals at m/z ) 65 [Cp+] to those at m/z ) 90 [Zr+] and m/z ) 155 [CpZr+] as a function of the reaction temperature. The intensities were measured during the Cp2Zr(CH3)2 pulse over a surface covered with -Cp2 - xZr(CH3)2 - y.
the decomposition of the metal precursor.12 The difference between decomposition temperatures seen in the actual thin film deposition studies and in the present QMS in situ studies is due to different reactor setups; the surface area in the present study is about 40 times larger and requires longer pulse lengths. Thus, even slow decomposition can be detected. A similar 50 °C difference has also been reported in the case of the titanium isopropoxide-water ALD process.18 The surface reaction byproducts released during the Cp2Zr(CH3)2/D2O ALD growth were CH3D (m/z ) 17) and CpD (m/z ) 67). Carbon dioxide (m/z ) 44) was not detected. At m/z ) 28, a slightly higher intensity peak was obtained during the growth cycles than from the metal precursor itself. This indicates release of ethene or carbon monoxide. However, the fragmentation of the zirconium precursor also caused a strong-intensity peak at m/z ) 28, and thus possible small amounts of species with that m/z value originating from the surface reactions would be difficult to detect. Figure 2 shows the QMS signals of the reaction byproducts CH3D and CpD obtained during two growth cycles of the Cp2Zr(CH3)2/D2O ALD process at 350 °C. The reaction byproducts were released during both the Cp2Zr(CH3)2 and D2O pulses. In ALD, the complete saturation of the substrate surface with the precursor pulses enables the self-limiting growth mode.6 The saturation was studied by doubling the Cp2Zr(CH3)2 precursor pulse length from 3 to 6 s and monitoring the total amount of reaction byproducts released during one complete ALD cycle. The dosage of the metal precursor had no significant effect on the total amount of CH3D or CpD released in the deposition temperature range of 200 to 400 °C (Figure 3). This indicates that complete surface saturation is already achieved with the 3 s metal precursor pulse length and the growth mode is of the ALD type. The CH3D is mainly released during the metal precursor pulse (Figure 4a). Only a relatively small amount (∼10%) of CH3D is released during the D2O pulse. At 210 °C, the amount of reaction byproducts was low. In the earlier study, it was reported that at 210 °C the growth rate was only 0.02 Å/cycle,12 which was proposed to be due to insufficient thermal energy available to promote the surface reactions. When the deposition temperature is increased, the amount of CH3D released during one ALD cycle increases, reaching its maximum at 350 °C. At this temperature, the surface reactions are fast enough, and the metal precursor is not yet decomposed. Thus, the growth temperature of 350 °C can be suggested as an
QMS Study of Atomic-Layer Deposition of ZrO2
Langmuir, Vol. 21, No. 16, 2005 7323
Figure 2. QMS signal obtained from (a) CH3D (m/z ) 17) and (b) CpD (m/z ) 67) during two ALD cycles at a reaction temperature of 350 °C. Pulse times are 3 s for both precursors and 5 s for the inert gas purge period. Figure 4. Amount of released (a) CH3D and (b) CpD during the Cp2Zr(CH3)2 or D2O pulse at different temperatures. The Cp2Zr(CH3)2 pulse time was 6 s, and the D2O pulse time was 3 s.
Figure 3. Total amount of reaction byproducts, CH3D (m/z ) 17) or CpD (m/z ) 67), released during one complete ALD cycle at different reaction temperatures and with a 3 or 6 s Cp2Zr(CH3)2 precursor pulse time. The D2O pulse time was 3 s.
optimal temperature for the Cp2Zr(CH3)2/H2O process. This is in good agreement with the compositional analysis: the stoichiometric ZrO2 films deposited at 350 °C contained less than 0.1 at. % C and H, but a slight increase in the impurity levels was observed if the deposition temperature was increased or decreased.12 During the metal precursor pulse, the amount of CpD (m/z ) 67) released stayed rather constant in the temperature range of 200-400 °C (Figure 4b). About 40% of the Cp ligands were released during the Cp2Zr(CH3)2 pulse. Above 400 °C, where the thermal decomposition of the precursor molecule is accelerated, the amount decreased. With increasing temperature, the density of -OD groups on the ZrO2 surface decreases,17 which may explain the lower amount of reaction byproducts CpD and also CH3D
(Figure 4a) during the Cp2Zr(CH3)2 pulse. However, during the D2O pulse the amount of CpD byproducts rapidly increases. When D2O is used as an oxygen source, the deuterated reaction byproducts can be formed only in the surface reactions with the -OD groups or in reactions with the arriving D2O molecules. The thermal decomposition of Cp2Zr(CH3)2 should not increase the amount of deuterated reaction byproducts. The increase of the amount of CpD byproducts during the D2O pulse at 440 °C must be due to a change in the reaction mechanism; Cp2Zr(CH3)2 chemisorbs mostly without exchange reactions whereas the density of -OD groups on the ZrO2 surface is lower at temperatures this high.17 The subsequent D2O pulse then releases the Cp and CH3 ligands. At the same time, the precursor decomposition plays a role in the reaction mechanism, and self-limiting ALD growth is not achieved. In the ALD growth of oxide thin films, the number of -OH groups left on the surface after the water pulse is the key factor in the controlled growth affecting the exchange reactions.6 Thus, it can be suggested that ZrO2 grows in the present Cp2Zr(CH3)2/D2O process via exchange reactions with the OD-terminated surface as the starting surface
(x + y)-OD(s) + Cp2Zr(CH3)2(g) f (-O-)(x + y)Cp2 - xZr(CH3)2 - y(s) + yCH3D(g) + xCpD(g) (1)
7324
Langmuir, Vol. 21, No. 16, 2005
Niinisto¨ et al.
(-O-)(x+y)Cp2 - xZr(CH3)2 - y(s) + 2D2O(g) f (-O-)2Zr(OD)4 - [(2 - x)+(2 - y)](s) + (2 - x)CpD(g) + (2 - y)CH3D(g) (2) During the first step of the ALD process, the metal precursor is introduced into the reaction chamber, and Cp2Zr(CH3)2 reacts with the surface -OD groups and releases the reaction byproducts (eq 1). For example, at 350 °C the amount of released CH3D during the metal pulse is about 90% of the overall amount released during one complete reaction cycle. Because the methyl ligand is expected to be extremely reactive, one could indeed assume that Cp2Zr(CH3)2 would react with the surface -OD groups via methyl ligands releasing all CH3D during the metal precursor pulse. In the present case, the amount of released CH3D is close to this mechanism, but, for example, at 350 °C CpD is released during both metal precursor and D2O pulses in a ratio of 3:5, respectively. In other words, about 40% of the CpD is released during the metal precursor pulse whereas the D2O pulse removes the rest of the ligands and converts the surface back to OD-terminated, ready for the next ALD cycle. Thus, it can be concluded that of the two ligands the methyls react more preferentially than the cyclopentadienyls. The proposed surface reactions of the current ALD process at 350 °C are presented in Figure 5. It should be noted that the scheme that is drawn is oversimplified and in 3D space the formation of bulk ZrO2 is a more complex. In the ideal but extremely rare case, one monolayer is formed during one ALD growth cycle. In the present case, the steric hindrances caused by the bulkiness of the cyclopentadienyl ligand significantly reduce the growth rate, and as a result, only a distinct fraction of a monolayer is formed during one growth cycle. Nevertheless, the suggested mechanism (cf. Figure 5) leads to stoichiometric (1:2) zirconium dioxide as confirmed by ion-beam analysis.12 In addition, because hydroxyl groups are the reactive sites in the ALD oxide growth,6 the density of these reaction sites has a major effect on the growth rate and reaction mechanism. The alternative reaction route, where the Zr-OD surface after one ALD cycle is dehydroxylated, is not likely at the present temperature because the Zr-OH surface species are rather stable against dehydroxylation.25 Figure 6 summarizes the effect of deposition temperature on the reaction mechanism. The percentage of CH3 or Cp ligands released during the metal precursor pulse stays rather constant at deposition temperatures ranging from 210 to 400 °C, meaning that the reaction mechanism does not significantly change. When comparing the current Cp-based ZrO2 process with the chloride-based process,17 it should be noted that the fraction of ligands released during the metal precursor pulse is slightly higher in the Cp-based process. In the ZrCl4-based process, about 4050% of the ligands were released during the metal precursor pulse at 250-375 °C,17 whereas in the present process the fraction was over 60%. In the Al(CH3)3/D2O process, the fraction was only slightly lower, viz. 50%, but decreased with increasing temperature.16 This reduction is caused by the decrease of hydroxyl groups, and it is even more strongly seen in the halide-based processes of HfO226 and TiO2.27 However, before the thermal decomposition of Cp2Zr(CH3)2 the decrease in the number (25) Agron, P. A.; Fuller, E. L.; Holmes, H. F. J. Colloid Interface Sci. 1975, 52, 553. (26) Aarik, J.; Aidla, A.; Kiisler, A.-A.; Uustare, T.; Sammelselg, V. Thin Solid Films 1999, 340, 110. (27) Matero, R.; Rahtu, A.; Ritala, M. Chem. Mater. 2001, 13, 4506.
Figure 5. Proposed surface reactions of Cp2Zr(CH3)2/H2O ALD process at 350 °C. The ALD cycle consists of the (a) metal precursor pulse on the OD-terminated surface, (b) metal precursor adsorbed and CH3D and CpD released, (c) inert gas purge and D2O pulse, and (d) CH3D and CpD released, the surface converted back to OD-terminated, and the inert gas purge, leaving the surface ready for the next ALD cycle.
Figure 6. Fraction of CH3 and Cp ligands released during the Cp2Zr(CH3)2 pulse with respect to the total amount released during one complete ALD cycle at different reaction temperatures.
hydroxyl groups does not seem to affect the reaction mechanism in the present study. As described, no major change in the reaction mechanism is observed until 440 °C. At this temperature, the thermal decomposition of the Cp2Zr(CH3)2 precursor destroys the self-limiting ALD growth mode. Conclusions Reactions in the ALD process of ZrO2 from Cp2Zr(CH3)2 and deuterated water were studied in situ by QMS. The reaction byproducts observed were CpD and CH3D. Almost
QMS Study of Atomic-Layer Deposition of ZrO2
all (90%) of the CH3D was released during the Cp2Zr(CH3)2 pulse in the exchange reactions with the surface OD groups at reaction temperatures of 210 to 400 °C. About 40% of the Cp ligands were released during the Cp2Zr(CH3)2 pulse. The subsequent D2O pulse released the remaining Cp ligands and converted the surface back to OD-terminated, ready for the next ALD cycle. The reaction mechanism is only weakly dependent on the deposition temperature before thermal decomposition of the Cp2Zr(CH3)2 starts and plays a role in the mechanism at deposition temperatures exceeding 400 °C.
Langmuir, Vol. 21, No. 16, 2005 7325
The cyclopentadienyl-type precursor in question, Cp2Zr(CH3)2, behaves as a promising ALD precursor, and organometallic Cp precursors in general may open new possibilities for the ALD of high-quality oxide thin films. Acknowledgment. Financial support from the Academy of Finland (projects 204742 and 205777) and for J.N. from the Jenny and Antti Wihuri Foundation is gratefully acknowledged. LA0500732
