pubs.acs.org/Langmuir © 2009 American Chemical Society
Reaction Mechanism Studies on Atomic Layer Deposition of Nb2O5 from Nb(OEt)5 and Water Kjell Knapas,* Antti Rahtu,† and Mikko Ritala Laboratory of Inorganic Chemistry, Department of Chemistry, University of Helsinki, P.O. Box 55, FIN-00014 University of Helsinki, Finland. †Current address: Vaisala OYJ, Vanha Nurmij€ arventie 21, 01670 Vantaa, Finland Received June 25, 2009. Revised Manuscript Received September 10, 2009 The reaction mechanism in the atomic layer deposition of Nb2O5 from Nb(OEt)5 and deuterated water was studied in situ with a quadrupole mass spectrometer (QMS) and a quartz crystal microbalance (QCM). The responses of these in situ measurement techniques to the characteristics of the ALD processes were thoroughly clarified and the process parameters carefully optimized. Also the effect of the reaction temperature on the extent of decomposition reactions of Nb(OEt)5 interfering with the ALD process was investigated. Decomposition did occur at 400 °C but not at temperatures of 350 °C and below. Also the reaction mechanism was studied as a function of reaction temperature and found to be about the same until decomposition of the precursor started. Deuterated ethanol was found to be the most important gaseous byproduct of the ALD process but also some diethyl ether apparently formed. About one of the five ethoxide ligands of Nb(OEt)5 was released during the Nb(OEt)5 pulse and the rest during the D2O pulse. Finally separate experiments were performed to study the adsorption of ethanol as well as 2-propanol on the surface of Nb2O5. Ethanol was found to adsorb. It could also be stated that water can replace the adsorbed ethanol. On the other hand 2-propanol apparently did not adsorb on the surface of Nb2O5.
Introduction Atomic layer deposition (ALD) is a sophisticated chemical gasphase method for depositing thin films.1-3 In ALD the precursors are pulsed one at a time into the reaction chamber and separated by purging with inert carrier gas. The growth mode is self-limiting since the precursors only saturate the surface with chemisorbed species. Therefore, one can accurately control the thickness of the films simply by adjusting the number of deposition cycles. The films also possess excellent uniformity and conformality. Usually ALD processes are studied with growth experiments, i.e. growing films and examining them afterward ex situ. Such an approach scarcely provides information on the actual surface reactions taking place. Understanding those reactions is crucial in controlling the processes and in developing new ones. A powerful means of getting knowledge of the ALD reactions is studying the processes in situ with a quadrupole mass spectrometer (QMS) and a quartz crystal microbalance (QCM).4 The quadrupole mass spectrometer follows the composition of the gas phase of the reaction chamber and the quartz crystal microbalance indicates the mass changes of the growing film. The release of gaseous byproducts of the deposition reactions and the mass changes of the quartz crystal during different precursor pulses give direct information on the reaction mechanisms. Nb2O5 is a high-permittivity material with potential thin film applications in for instance DRAM devices (dynamic random access memory), optics, and gas sensors.5 Original growth experi*To whom correspondence should be addressed. E-mail: Kjell.Knapas@ helsinki.fi. (1) Suntola, T.; Antson, J. U.S. Patent 4058430, 1977. (2) Ritala, M.; Leskel€a, M. In Handbook of Thin Film Materials; Nalva, H. S., Ed.; Academic Press: San Diego, CA, 2001: Vol. 1, Chapter 1. (3) Leskel€a, M.; Ritala, M. Angew. Chem., Int. Ed. 2003, 42, 5548. (4) Knapas, K.; Ritala, M. Chem. Mater. 2008, 20, 5698 and references therein. (5) Knapas, K.; Rahtu, A.; Ritala, M. Chem. Vap. Deposition, in press, and references therein.
848 DOI: 10.1021/la902289h
ments to deposit Nb2O5 using the NbCl5-H2O ALD process failed due to etching of Nb2O5 by NbCl5.6 Recently we demonstrated that NbCl5 and H2O may still be applicable for depositing Nb2O5 by ALD if some Nb2O5 is allowed to react with NbCl5 before the substrates.5 In this approach the niobium containing species giving rise to growth of Nb2O5 is NbOCl3. However this technique may be somewhat difficult to apply in practice. Fortunately there is the Nb(OEt)5-H2O process that demonstrates almost ideal ALD behavior.7 That feature also makes this particular process suitable for a thorough mechanism analysis, which is presented here. Preliminary studies on the reactions of this process and also of the related Ta(OEt)5-H2O process were made in the QMS study of the Ti(OEt)4-H2O process previously conducted by our group8 and before that also in the article where we originally described integration of the QMS into an ALD reactor.9
Experimental Section The experiments were carried out in a specially modified10 commercial flow-type ALD reactor manufactured by ASM Microchemistry Ltd. The pressure of the reactor was 3 mbar and the area of the soda lime glass substrates 3500 cm2. Argon (Oy AGA Ab, 99.999%) was used as the carrier gas. The QMS was a Hiden HAL/3F 501 RC and a Faraday detector and an ionization energy of 70 eV were used. The QMS chamber was pumped differentially through a 200 μm orifice giving a pressure of 1 10-4 mbar, except for the alcohol adsorption measurements, where a 50 μm orifice was used and (6) Elers, K.-E.; Ritala, M.; Leskel€a, M.; Rauhala, E. Appl. Surf. Sci. 1994, 82/83, 468. (7) Kukli, K.; Ritala, M.; Leskel€a, M.; Lappalainen, R. Chem. Vap. Deposition 1998, 4, 29. (8) Rahtu, A.; Kukli, K.; Ritala, M. Chem. Mater. 2001, 13, 817. (9) Ritala, M.; Juppo, M.; Kukli, K.; Rahtu, A.; Leskel€a, M. J. Phys. IV 1999, 9, Pr8-1021. (10) Rahtu, A.; Ritala, M. Electrochem. Soc. Proc. 2000, 2000-13, 105.
Published on Web 10/01/2009
Langmuir 2010, 26(2), 848–853
Knapas et al.
Article
led to a pressure of 3 10-6 mbar. The QCM was a modified Maxtek TM 400 with a sampling rate of 20 Hz. Nb(OEt)5 (Inorgtech) was held inside the reactor in an open boat at 100 °C and pulsed with inert gas valving.2 D2O (Eurisotop, 99.9%) was used instead of H2O to better distinguish the reaction byproducts from species forming in fragmentation and recombination reactions in the QMS ionizator. Still weak background signals were detected at the m/z values corresponding to the deuterated products when the same precursor was pulsed repeatedly, so these were subtracted from the signals measured during the ALD reactions.8 D2O, ethanol (Primalco ETAX, 94.0%) and 2-propanol (Fisher, 99.99%) were held in bottles outside the reactor at room temperature and led into the reactor through solenoid valves. A precursor pulse and the following purge period are denoted “a/b precursor”, where a is the length of the precursor pulse [s] and b the length of the purge period [s]. The ALD process cycle consists of two successive precursor pulses and purge periods and is denoted “a1/b1-a2/b2 Nb(OEt)5, D2O”. Precursor and process cycle sequences are denoted n[a/b] and n[a1/b1-a2/b2] which mean that a/b or a1/b1-a2/b2 is repeated n times. Each experiment consists of several of these kinds of sequences. When investigating the Nb(OEt)5-D2O process the following pulsing sequence was used: 1[10/10]D2O þ 5[y/(xþz1þz2)]D2O þ 10[x/z1-y/z2]Nb(OEt)5, D2O þ 3[x/(yþz1þz2)]Nb(OEt)5. The aim of the first water pulse is to ensure that the surface is fully hydroxyl terminated. The following water pulses are reference pulses which give the water background signals for the m/z values being measured. Accordingly the Nb(OEt)5 pulses at the end are reference pulses that give the Nb(OEt)5 backgrounds. Of course, during the first of these pulses, exchange reactions are still taking place. However the following two should give equal signals. If they do not, surface saturation is not achieved during one pulse and the parameters are outside ALD conditions. When studying the effect of the length of the Nb(OEt)5 pulse x was 1, 3, 5, and 7, and y, z1, and z2 were 3. When studying the effect of the length of the D2O pulse x was 5, y was 1, 3, 5, and 7, and z1 and z2 were 3. When studying the effect of the length of the purge periods x was 5, y was 3 and z1 and z2 were 1, 3, 5, and 7, or z1 was 0.5, 1.5, and 3 and z2 was 3, or z1 was 3 and z2 was 0.5, 1.5, and 3. When studying the effect of reaction temperature, x was 5 and y, z1 and z2 were 3 and reaction temperatures were 150, 200, 225, 250, 275, 300, 325, 350, and 400 °C. When studying the decomposition of Nb(OEt)5 the pulsing sequence 5[1/10]Nb(OEt)5 was used. When studying the adsorption of alcohols on the surface of Nb2O5 the pulsing sequence was 5[0.5/10]D2O þ 10[5/3-3/3]Nb(OEt)5, D2O þ 1[4/y]EtOH or i PrOH þ 5[0.5/10]D2O. y was 80 with EtOH and 20 with iPrOH. In these experiments, the reaction temperature was 250 °C.
Results and Discussion ALD Process. The Nb(OEt)5-D2O ALD process deposits thin films of Nb2O5 predominantly according to the net reaction 1 where the gaseous byproduct is deuterated ethanol, EtOD. It is to be noted that the process does not proceed fully in this course but produces also at least small amounts of diethyl ether as shall later be seen. However these other reactions are minor as compared to reaction 1. 2NbðOEtÞ5 ðgÞ þ 5D2 OðgÞ f Nb2 O5 ðsÞ þ 10EtODðgÞ
ð1Þ
Figure 1 shows the QCM pattern of three cycles of the ALD process at 225 °C. The mass increases by the extent m1 during the Nb(OEt)5 pulse and decreases during the D2O pulse. The decrease is however not as large as the increase, so the mass increases by the extent m0 in netto during each cycle. The mass increase starts immediately when the Nb(OEt)5 pulse reaches the crystal and stops when the pulse ends. The increase is however not even but it Langmuir 2010, 26(2), 848–853
Figure 1. QCM data in the Nb(OEt)5-D2O process at 225 °C. The bold horizontal bars indicate the timings of the Nb(OEt)5 and D2O pulses, respectively.
Figure 2. Comparison of QCM patterns in the Nb(OEt)5-D2O process at 225 °C with 1, 3, 5, and 7 s long Nb(OEt)5 pulses.
almost stops quite quickly after the beginning of the pulse and continues after a while. The mass decrease starts immediately when the D2O pulse reaches the crystal, slows down exponentially and has almost stopped when the pulse ends. In Figure 2 the QCM patterns during one ALD cycle at 225 °C with 1, 3, 5, and 7 s long Nb(OEt)5 pulses are compared. It can be seen that regardless of the length of the Nb(OEt)5 pulse the mass increase stops temporarily at about the same level. With 1 s long Nb(OEt)5 pulses the mass does not increase anymore from this level but with longer pulses the mass increase continues after about 2 s. Irrespective of the pulse length the mass stays constant during the purge. Furthermore, one can see that m0 increases with increasing pulse length but this increase gets smaller the longer the Nb(OEt)5 pulse is. The QMS results of EtOD (m/z = 47) during three ALD cycles at 225 °C are displayed in Figure 3. Much more EtOD is released during the D2O pulse than during the Nb(OEt)5 pulse. During the Nb(OEt)5 pulse the intensity reaches a constant level quite fast and returns evenly to the original level during the purge. During the D2O pulse the intensity rises at first almost vertically and starts then to fall exponentially. When the pulse ends the fall accelerates for a while. The growth rate of the film was investigated at all reaction temperatures as a function of the lengths of the precursor pulses and the purge periods. In the QCM measurements the growth rate is depicted by the mass increase m0 (cf. Figure 1). This quantity was determined as the average from the nine last ALD cycles in each measurement. Furthermore, averages were calculated when DOI: 10.1021/la902289h
849
Article
Knapas et al.
Figure 3. QMS data (m/z = 47) in the Nb(OEt)5-D2O process at
225 °C.
several measurements were done using the same parameters. In the QMS measurements the growth rate due to exchange reactions is obtained with m/z = 47 (EtOD). From each measurement only the last ALD cycle and, for the background, the last precursor pulses on the surface saturated with the same precursor have been taken into account. The intensities of these peaks were integrated with time and the corresponding background intensities were subtracted from the intensities during the ALD exchange reactions. The sum of the intensities thus obtained for the different precursors depicts the growth rate and is noted Σp. From these values averages have been calculated when several measurements have been performed using the same parameters. In Figures 4 a-b the measurements as a function of the length of the Nb(OEt)5 pulse at 225 °C are displayed. According to both the QCM and the QMS results the growth rate saturates with 5 s Nb(OEt)5 pulses. This was principally found to be the case at all the other reaction temperatures as well. Accordingly the growth rate was found to saturate with 3 s D2O pulses at all the reaction temperatures. On the other hand the lengths of the purge periods did not seem to affect the growth rate. In Figure 5, the growth rate according to both QCM (m0) and QMS (Σp) is shown as a function of reaction temperature. One can say that QCM and QMS match quite well with each other. The growth rate would seem to first increase rapidly when the temperature rises from 150 to 200 °C and then to decrease when the temperature increases to 225 °C. A peak this sharp is not very likely, so it is possible that there is some inconsistency in the measurement at 200 or 225 °C. Generally one can say that the growth rate increases in the reaction temperature interval 150-300 °C and decreases between 300 and 400 °C. It appears however that both the increase and the decrease include some abnormalities. According to these measurements the optimum reaction temperature giving the highest growth rate is between 275 and 300 °C which is slightly higher than between 230 and 260 °C which was obtained with the growth experiments.7 Decomposition of Nb(OEt)5. The main fragment of Nb(OEt)5 (Nb(OEt)4þ, m/z = 273) was not detected at all at the highest reaction temperature of 400 °C, whereas its intensity was about the same at all the other reaction temperatures investigated. During the ALD process the intensity of Nb(OEt)4þ is a bit smaller than during the reference Nb(OEt)5 pulses when Nb(OEt)5 is pulsed several times running. This is of course because Nb(OEt)5 is partially consumed in the ALD process. From these observations one can conclude that Nb(OEt)5 decomposes at 400 °C. The most likely decomposition reactions produce solid Nb2O5 and ethanol and ethene, reaction 2, or diethyl ether, 850 DOI: 10.1021/la902289h
Figure 4. (a) QCM and (b) QMS results (m/z = 47) in the
Nb(OEt)5-D2O process at 225 °C as functions of Nb(OEt)5 pulse length. The symbol p means the intensity associated with the precursor Nb(OEt)5 and Σp means the sum of the intensities associated with the precursors Nb(OEt)5 and D2O. The meanings of m0 and m1 are apparent in Figure 1.
Figure 5. The QCM and QMS results depicting the growth rate of the Nb(OEt)5-D2O process as functions of reaction temperature.
reaction 3. 2NbðOEtÞ5 ðgÞ f Nb2 O5 ðsÞ þ 5EtOHðgÞ þ 5C2 H4 ðgÞ 2NbðOEtÞ5 ðgÞ f Nb2 O5 ðsÞ þ 5Et2 OðgÞ
ð2Þ ð3Þ
In the decomposition reaction 2, the main fragment of ethanol is CH2OHþ, m/z = 31, and the molecular peak of ethene m/z = 28. These ions are formed also directly as fragments of Nb(OEt)5 in the QMS. The intensities of these m/z values followed the variations in the intensity of m/z = 273 (Nb(OEt)4þ) otherwise, Langmuir 2010, 26(2), 848–853
Knapas et al.
Figure 6. The intensities of m/z = 28 and m/z = 273 and the ratio of the intensity of m/z = 28 to the sum of the intensities of m/z = 28 and m/z = 273 in the Nb(OEt)5-D2O process as functions of reaction temperature.
Figure 7. Results of m/z = 74 in the Nb(OEt)5-D2O process at different reaction temperatures.
but they did not disappear at the reaction temperature of 400 °C, but stayed about the same. In Figure 6 the intensity of m/z = 28 is shown together with the simultaneously measured intensity of m/z = 273. Also the ratio of the intensity of m/z = 28 to the sum of the intensities of m/z = 28 and m/z = 273 is depicted and, as it can be seen, the variations in this ratio are much smaller than in the intensities themselves. M/z = 31 behaves similar to m/z = 28. In the measurements Nb(OEt)5 was pulsed several times running and only the last pulses were taken into account and integrated with time. The results support very well the conclusion that Nb(OEt)5 decomposes at 400 °C but not at the lower temperatures investigated. The fact that the intensities of m/z = 28 and m/z = 31 remain unchanged also when going to 400 °C can be understood as follows: the decomposition products magnify the intensities of these m/z values, but there is less, if any, Nb(OEt)5 to fragment, which diminishes the intensities correspondingly. The decomposition reaction 3 produces diethyl ether with the molecular peak at m/z = 74. Figure 7 shows one measurement of this m/z value from all the investigated reaction temperatures. Peaks rise during the Nb(OEt)5 pulses both during the ten ALD cycles and during the three successive metal precursor reference pulses at the end of the measurements. The intensities of the peaks increase exponentially with the reaction temperature. In Figure 8 the intensity of m/z = 74 is depicted as a function of the length of the Nb(OEt)5 pulse at 400 °C and a linear increase is observed. Also, the intensity is independent of with which precursor, Nb(OEt)5 or D2O, the surface was saturated before the pulse (Figure 7). Here the conclusion is that diethyl ether at this Langmuir 2010, 26(2), 848–853
Article
Figure 8. Intensity of m/z = 74 in the Nb(OEt)5-D2O process at 400 °C as a function of Nb(OEt)5 pulse length.
temperature (400 °C) is originating from decomposition of Nb(OEt)5. At the lower temperatures the intensity of m/z = 74 was not observed to increase linearly with Nb(OEt)5 pulse length but to saturate with 5 s pulses. Therefore, the diethyl ether cannot there be at least entirely due to decomposition reactions, but it must have other origins as well. It appears that diethyl ether forms in small amounts in the ALD exchange reactions, since the intensity of m/z = 74 is greater during the ALD process than when Nb(OEt)5 is pulsed several times successively. This is quite possible since it only requires a combination reaction of two ethoxy groups on the surface, a reaction that is known to occur very well on metal oxide surfaces11,12 and especially also on the niobium oxide surface.13 Furthermore, this m/z value must form in the ionizer of the QMS as well. Only at 400 °C are additionally observed weak signals at m/z = 58 and m/z = 60. The intensities increase approximately linearly with the Nb(OEt)5 pulse length and are independent of with which precursor the surface was saturated before the pulse. Therefore, they also arise from decomposition products. m/z = 58 is the molecular peak of C4H10 (butane) and m/z = 60 the molecular peak of CH3COOH (acetic acid). Though such a reaction is rarely reported, on a surface of chromium(III)oxide methanol formed ethane through dimethyl ether.14 Maybe butane is formed here accordingly through diethyl ether. Formation of corresponding carboxylate surface species from alcohols is also very well documented on metal oxide surfaces,15,16 so the release of the corresponding carboxylic acid is not that big a surprise here either. Anyway behaviors of m/z = 74, m/z = 58, and m/z = 60 support very well the conclusion that Nb(OEt)5 decomposes at 400 °C but not at lower temperatures. This is in fairly good agreement with decomposition studies of titanium alkoxides.17 ALD Reaction Mechanism. When following the net reaction 1 the Nb(OEt)5-D2O ALD process would proceed through the half-reactions 4 and 5. The arriving Nb(OEt)5 meets some -OD groups on the surface causing thereby partial release of the ethoxy ligands in deuterated form. The following D2O pulse releases the rest of the ligands as EtOD and restores the -OD termination of the surface. In the reaction equations n denotes the number of EtOD molecules released during the Nb(OEt)5 pulse per (11) DeCanio, E. C.; Nero, V. P.; Bruno, J. W. J. Catal. 1992, 135, 444. (12) Lusvardi, V. S.; Barteau, M. A.; Farneth, W. E. J. Catal. 1995, 153, 41. (13) DeCanio, E. C.; Nero, V. P.; Ko, E. I. J. Catal. 1994, 146, 317. (14) Yamashita, K.; Naito, S.; Tamaru, K. J. Catal. 1985, 94, 353. (15) Hertl, W.; Cuenca, A. M. J. Phys. Chem. 1973, 77, 1120. (16) Hussein, G. A. M.; Sheppard, N.; Zaki, M. I.; Fahim, R. B. J. Chem. Soc. Faraday Trans. 1 1991, 87, 2661. (17) Rahtu, A. Atomic Layer Deposition of High Permittivity Oxides: Film Growth and In Situ Studies; Ph.D. Thesis; University of Helsinki: 2002; pp 43-44, thesis available at http://ethesis.helsinki.fi.
DOI: 10.1021/la902289h
851
Article
Knapas et al.
form diethyl ether, and the occurring reaction may well be catalyzed by the Nb2O5 surface and can therefore still uphold a self-limiting growth mechanism. Reaction 8, formation of an ether molecule from two alkoxy groups on metal oxide surfaces is a well-known reaction.11,12 Experiments with single crystal surfaces have demonstrated that the alkoxy groups evidently have to be bonded to the same metal atom.18 Also, the corresponding reaction between an alkoxy group on the surface and an alcohol molecule in the gas phase or a physisorbed alcohol molecule does not occur, since a surface treated with CH3OH (CD3OD) forms almost exclusively CH3OCH3 (CD3OCD3) even if CD3OD (CH3OH) is present in the gas phase during the reaction.19 Figure 9. QCM and QMS results depicting the reaction mechan-
0:41-ODðsÞ þ NbðOEtÞ5 ðgÞ
ism of the Nb(OEt)5-D2O process as functions of reaction temperature. The notation n means the number of EtOD molecules released from the surface during the Nb(OEt)5 pulse per Nb(OEt)5 molecule.
f ð -OÞ0:85 NbðOEtÞ3:71 ðsÞ þ 0:44Et2 OðgÞ þ 0:41EtODðgÞ
Nb(OEt)5 molecule. From the QMS results n is obtained by dividing the amount of EtOD (m/z = 47) released during the Nb(OEt)5 pulse by the sum of the amounts of EtOD released during the Nb(OEt)5 and D2O pulses and multiplying by five, the number of ethoxy groups in one Nb(OEt)5 molecule. From the QCM results n is deduced with the help of the ratio of m0 and m1, which were defined above. As calculated by molecular masses, in half-reaction 4, m1 = M(Nb(OEt)5) - nM(EtOD) = 318-47n and of course m0 = M(NbO2.5) = 133. Therefore, m0/m1 = 133/(318-47n) and n = [318-133/(m0/m1)]/47.
f ð -OÞ2:5 NbðODÞ0:41 ðsÞ þ 3:71EtODðgÞ
ð7Þ
-M ðORÞx ðsÞ f -MOðORÞx -2 ðsÞ þ R2 OðgÞ
ð8Þ
n-ODðsÞ þ NbðOEtÞ5 ðgÞ f ð -OÞn NbðOEtÞ5 -n ðsÞ þ nEtODðgÞ
ð4Þ
ð -OÞn NbðOEtÞ5 -n ðsÞ þ 2:5D2 OðgÞ f ð -OÞ2:5 NbðODÞn ðsÞ þ ð5 -nÞEtODðgÞ
ð5Þ
The results for n in reactions 4 and 5 according to QCM and QMS are displayed in Figure 9 as functions of reaction temperature. According to the QCM, n is about 2.5 at 400 °C and on average about 1 at lower reaction temperatures. According to the QMS, n is about 1 at 400 °C and about 0.5 at lower temperatures. Since Nb(OEt)5 decomposes at 400 °C, differing results with QCM and QMS are no surprise there. Actually the results can be very well understood on the basis of the decomposition, since they show that according to QCM the mass increases during the Nb(OEt)5 pulse more than it should according to QMS. At reaction temperatures lower than 400 °C the results obtained with QCM and QMS are much more consistent. The small differences can be at least partially explained with the fact that some diethyl ether appears to form as an exchange reaction byproduct during the Nb(OEt)5 pulse as explained above and considered in our preliminary study on this process.8,9 Therefore, more ligands are released from the surface during the Nb(OEt)5 pulse than the intensity of m/z = 47 implies. Consequently n obtained from m/z = 47 becomes a bit too small. In fact, from the QCM results and the results from m/z = 47 one can deduce also the amount of diethyl ether formed if that is supposed to explain the difference. Reactions 6 and 7 represent a fit to the results where n as obtained above was 1.0 according to QCM and 0.5 according to QMS. In these reactions part of the oxygen atoms of Nb2O5 film come from the ethoxy groups of Nb(OEt)5. Though such a net reaction resembles the decomposition reaction 3, all the ethoxy groups in the reacting Nb(OEt)5 molecules do not need to 852 DOI: 10.1021/la902289h
ð6Þ ð -OÞ0:85 NbðOEtÞ3:71 ðsÞ þ 2:06D2 OðgÞ
Reaction 6 also provides a plausible explanation for the two distinct mass growth regimes observed in the QCM pattern during the Nb(OEt)5 pulse (Figure 1) and described above. One of these regimes may be due to the reaction path producing EtOD, and the other due to the path producing Et2O. It is reasonable to assume that at first the surface -OD groups react with the adsorbing Nb(OEt)5 molecules producing EtOD and -Nb(OEt)x surface species and therefore causing the first mass increase. Then some of the -OEt groups on the surface react producing Et2O (reaction 8) and opening sites for molecular adsorption of some additional Nb(OEt)5 molecules, which corresponds to the second mass increase. One could still argue that it is possible that some EtOD released during the Nb(OEt)5 pulse is readsorbed on the surface, reaction 9, and is then replaced with water during the water pulse, reaction 10, since the amount of EtOD observed during the Nb(OEt)5 pulse is surprisingly small. Adsorption of alcohols on metal oxide surfaces is an extensively studied subject and produces many surface species: the alcohol molecules may bond intact through hydrogen and coordination bonds, and form alkoxy group-hydroxyl group pairs through dissociative chemisorption.11,12 However, the sequence depicted in reactions 9 and 10 cannot be distinguished from the measurements on the ALD process where only the final release of molecules from the surface is detected. Below separate experiments verify however that EtOH does chemisorb on the surface of Nb2O5 and is released by D2O. Still this does not necessarily mean that readsorption would occur during the ALD process, since the partial pressure of ethanol is in any case much smaller there than in the separate experiments below. However, if readsorption occurs during the ALD process, this could in part explain the low growth rate observed with the process,7 since alkoxy groups or coordinated alcohol molecules may block hydroxyl groups from incoming Nb(OEt)5 molecules. Therefore, in this case the low growth rate does not necessarily mean that the hydroxyl group concentration is exceptionally small. EtODðgÞ f EtODðadsÞ EtODðadsÞ þ D2 OðgÞ f D2 OðadsÞ þ EtODðgÞ
ð9Þ ð10Þ
(18) Kim, K. S.; Barteau, M. A. Surf. Sci. 1989, 223, 13. (19) Matsushima, T.; White, J. M. J. Catal. 1976, 44, 183.
Langmuir 2010, 26(2), 848–853
Knapas et al.
Article
sharply. This verifies that EtOD adsorbed on the surface and was released by D2O which has a smaller molecular mass. QMS verifies that EtOH is released during the first D2O pulse, so it was indeed adsorbed on the surface. In the experiments with 2-propanol, no adsorption and release was observed. The QCM signal returned to its original level shortly after the 2-propanol pulse, and no change was seen during the following D2O pulses. Also, no 2-propanol was detected with QMS during the D2O pulses succeeding the 2-propanol pulse. The conclusion is that ethanol can adsorb on the surface of Nb2O5 and be released with water, but 2-propanol cannot adsorb on the surface. On the basis of these findings, it would be interesting to compare the Nb(OEt)5-D2O process with the Nb(OiPr)5-D2O process, but this one remains still to be developed. In the case of TiO2 ALD, however, there is a marked difference between the Ti(OEt)4-D2O and Ti(OiPr)4-D2O processes: in the former only about 10% of the only byproduct ROD is released during the titanium alkoxide pulse as compared to a complete ALD cycle,8 but in the latter the corresponding ratio is about 50%.20
Conclusions
Figure 10. (a) QCM and (b) QMS data (m/z = 46) depicting adsorption of EtOH on the surface of Nb2O5. After 10 ALD cycles depositing Nb2O5 4/80 EtOH and 5 0.5/10 D2O follow.
Adsorption of Alcohols on the Surface of the Nb2O5 Thin Films. Adsorption of ethanol and 2-propanol (isopropanol) on the surface of Nb2O5 thin films grown from Nb(OEt)5 was studied with separate experiments at 250 °C. Parts a and b of Figure 10 show the QCM and QMS (m/z = 46, EtOH) results of one such measurement. According to QCM, the mass increases during the ethanol pulse. During the following purge period the signal decreases but does not return to the level before the ethanol pulse. Then the signal increases again when the purge still proceeds. During the first following D2O pulse the mass decreases
Langmuir 2010, 26(2), 848–853
In light of the experimental evidence obtained here with QCM and QMS, ALD of Nb2O5 from Nb(OEt)5 and D2O appears to proceed without interference from decomposition reactions of Nb(OEt)5 at reaction temperatures of 350 °C and below, the optimum temperature being between 275 and 300 °C. Though some Et2O apparently forms in the reactions besides the major byproduct EtOD. The reaction producing Et2O corresponds formally to a decomposition reaction but it can be surface catalyzed at temperatures below 400 °C where the decomposition starts. Therefore, it may still uphold pure ALD with a self-limiting growth mechanism. The amounts of both Et2O and EtOD released during the Nb(OEt)5 pulse are anyway quite small, and together they stand for only about one of the five ligands of Nb(OEt)5, the other four being released as EtOD during the water pulse. Acknowledgment. The Academy of Finland is acknowledged for funding this work. (20) Rahtu, A.; Ritala, M. Chem. Vap. Deposition 2002, 8, 21.
DOI: 10.1021/la902289h
853
