Characterization of titanium dioxide atomic layer growth from titanium ethoxide and water

Thin Solid Films 370 (2000) 163±172 www.elsevier.com/locate/tsf Characterization of titanium dioxide atomic layer growth from titanium ethoxide and ...
Author: Angelica Farmer
11 downloads 0 Views 389KB Size
Thin Solid Films 370 (2000) 163±172

www.elsevier.com/locate/tsf

Characterization of titanium dioxide atomic layer growth from titanium ethoxide and water Jaan Aarik a,*, Aleks Aidla a, VaÈino Sammelselg b, Teet Uustare a, Mikko Ritala c, Markku LeskelaÈ c È likooli Street, 50090 Tartu, Estonia Institute of Materials Science, University of Tartu, 18 U b Institute of Physics, University of Tartu, 142 Riia Street, 51014 Tartu, Estonia c Department of Chemistry, University of Helsinki, P.O. Box 55 (A. I. Virtasen aukio 1) FIN 00014 University of Helsinki, Finland a

Received 9 November 1999; received in revised form 10 February 2000; accepted 23 February 2000

Abstract Atomic layer growth of titanium dioxide from titanium ethoxide and water was studied. Real-time quartz crystal microbalance measurements revealed that adsorption of titanium ethoxide is a self-limited process at substrate temperatures 100±2508C. A relatively small amount of precursor ligands was released during titanium ethoxide adsorption while most of them was exchanged during the following water pulse. At temperatures 100±1508C, incomplete reaction between surface intermediates and water hindered the ®lm growth. Nevertheless, the deposition rate reached 0.06 nm per cycle at optimized precursor doses. At substrate temperatures above 2508C, the thermal decomposition of titanium ethoxide markedly in¯uenced the growth process. The growth rate increased with the reactor temperature and titanium ethoxide pulse time but it insigni®cantly depended on the titanium ethoxide pressure. Therefore reproducible deposition of thin ®lms with uniform thickness was still possible at substrate temperatures up to 3508C. The ®lms grown at 100±1508C were amorphous while those grown at 1808C and higher substrate temperature, contained polycrystalline anatase. The refractive index of polycrystalline ®lms reached 2.5 at the wavelength 580 nm. q 2000 Elsevier Science S.A. All rights reserved. Keywords: Atomic layer deposition; Titanium dioxide; Growth mechanism; Surface roughness

1. Introduction Titanium dioxide (TiO2) thin ®lms and dielectric structures containing TiO2 as a component have a number of perspective applications in microelectronics and sensor technology. In many applications, the thickness of these ®lms and structures as well as the thickness uniformity should be very precisely controlled. Atomic layer deposition (ALD), also known as atomic layer epitaxy [1] and molecular layering [2], is a method which allows this kind of control with a submonolayer accuracy and enables one to get uniform thickness even on pro®led substrates. This is due to the cyclic self-controlled character of the ALD process that results in a constant thickness increase in each deposition cycle. As the thickness increase per cycle is usually less than one monolayer of deposited substance and, in the ideal case, it does not depend on modest variations of process parameters, the thickness control is convenient and accurate. Several titanium compounds have been applied as precur* Corresponding author. Tel.: 1 372-7-375-877; fax: 1 372-7-375-540.

sors in atomic layer deposition of TiO2 thin ®lms [2±13]. Titanium chloride (TiCl4) is the precursor most frequently used together with water [2±9] or hydrogen peroxide [10] for ALD growth at substrate temperatures (Ts) ranging from 27 to 6008C. It has been shown that this precursor enables one to grow high-quality thin ®lms. For instance refractive index as high as 2.6 [3,8] and optical losses below 100 cm 21 [5] have been obtained for the ®lms grown by ALD from TiCl4 and water. Moreover, using TiCl4 as the titanium precursor, thin ®lms with different crystal structure and physical properties can be obtained [3±6,8]. Unfortunately, TiCl4 is a highly corrosive material. It has to be handled with care and special measures should be applied for protection of vacuum and gas lines. Also, signi®cant chlorine contamination has been observed in the ®lms grown at substrate temperatures close to 1008C [5]. In addition, HCl released in the growth process may, because of its high reactivity, readsorb and cause appearance of thickness pro®les in the gas ¯ow direction [14]. Therefore alternative titanium sources for TiO2 atomic layer deposition are of practical interest. Titanium isopropoxide (Ti(OCH(CH3)2)4) [11,12] and titanium ethoxide (Ti(OCH2CH3)4) [13] combined with

0040-6090/00/$ - see front matter q 2000 Elsevier Science S.A. All rights reserved. PII: S 0040-609 0(00)00911-1

164

J. Aarik et al. / Thin Solid Films 370 (2000) 163±172

water have also been used as precursors for atomic layer deposition of TiO2. The ALD growth has been performed in the substrate temperature ranges of 50±350 and 200±4008C for titanium isopropoxide [11,12] and titanium ethoxide [13], respectively. Although thin ®lms of uniform thickness have been obtained from these precursors, the growth rate signi®cantly depended on the substrate temperature as well as on the precursor doses [12,13]. A disadvantage of titanium ethoxide has been a rather low growth rate (0.03±0.04 nm/cycle) at substrate temperatures below 3008C. For instance, at substrate temperatures of 100±1508C, TiO2 growth rates exceeding 0.07 nm per cycle [5] have been obtained using titanium chloride and water as precursors. Different growth rate could be due to the larger size of the titanium ethoxide molecule compared with that of titanium chloride [13]. However, some other reasons, such as incomplete exchange reactions during precursor pulses and low density of activated adsorption sites, can also be proposed to explain slow thin ®lm growth. Therefore kinetic studies are inevitable to understand the reaction mechanisms and processes limiting the deposition rate. In the present study we investigated the kinetics of atomic layer deposition of titanium dioxide thin ®lms from titanium ethoxide and water by using the quartz crystal microbalance (QCM) method [5,6,15]. Besides real-time measurement of the mass increase per complete ALD cycle, the method enabled us to record the deposition kinetics within a single ALD cycle. In this way we obtained valuable information providing a possibility to draw some conclusions about surface reactions and optimize the conditions for thin ®lm growth. 2. Experimental A hot-wall low-pressure (250 Pa) ALD reactor equipped with a quartz crystal microbalance (QCM) [15] was used for growing the thin ®lms as well as for the real-time measurements. Titanium dioxide thin ®lms were deposited from alternating titanium ethoxide and water vapor pulses led into the reaction zone with N2 carrier gas of 99.999% purity. In order to remove the gaseous reaction products and to avoid overlapping of precursor pulses, the reactor was purged with pure carrier gas after each precursor pulse. The vapor pressure of titanium ethoxide was determined by the temperature of the titanium ethoxide cell varied from 48±878C during the QCM measurements and kept at 788C while growing the thin ®lms for ex situ studies. Water vapor was generated in a container kept at 208C whereas the vapor ¯ow into the reactor was controlled with a calibrated needle valve. In the real-time QCM experiments, the thin ®lm was grown directly on the mass sensor. However, before starting the real measurements a 2±3 nm thick TiO2 buffer-layer was always grown on the sensor. This allowed us to obtain data that described the reactions on the surface of TiO2 and did

not depend on the characteristics of the bare mass sensor surface. In order to reduce the experimental error and avoid misleading effects of transient processes that usually appeared during the ®rst ALD cycles applied after changing deposition parameters, the mass sensor signal was recorded for 3±5 subsequent cycles. Then the data corresponding to the ®rst cycle were neglected and the average response describing the mass changes during one ALD cycle was calculated from the rest of data. The thin ®lms for post-growth characterization were deposited on fused silica and (100)-oriented silicon substrates. During the thin ®lm growth the substrate surface was parallel to the gas ¯ow direction. The substrate temperature ranged from 100 to 3508C. The composition of the ®lms was determined with the electron probe microanalysis (EPMA) [16] and Auger electron spectroscopy (AES) methods. X-ray diffraction (XRD) and re¯ection high energy electron diffraction (RHEED) methods were used to study the phase composition and crystallinity. The optical thickness, refractive index, and extinction coef®cient were determined from transmission spectra using the method proposed by Swanepoel [17]. Atomic force microscopy (AFM) was used to investigate the surface morphology.

3. Results and discussion 3.1. Deposition kinetics In order to characterize adsorption of precursors, the time dependence of the thin ®lm mass was ®rst studied at relatively long cycle times. The curves recorded at the substrate temperatures of 215 and 3508C during a single ALD cycle are depicted in Fig. 1a. In both cases the titanium ethoxide pulse, ®rst purge time, H2O pulse and second purge time were 20, 10, 10 and 10 s, respectively. As can be seen in Fig. 1a, the titanium ethoxide pulse causes a signi®cant increase in the deposit mass. However, the mass starts to increase after a delay only. This kind of behavior indicates that a well-de®ned adsorption wave is formed in the reactor. As the adsorption wave propagates with a ®nite rate from the reactor inlet towards the mass sensor, the response does not appear immediately after switching on the reactant pulse. The delay time depends on the adsorption capacity of the reactor walls, on the concentration of precursor molecules in the gas phase and on the ¯ow rate of the carrier gas [18]. The adsorption of titanium ethoxide rather well saturates at the substrate temperature of 2158C (Fig. 1a). Thus, a surface intermediate layer which does not adsorb additional titanium ethoxide is formed at this temperature. At 3508C, however, the behavior of the mass sensor signal is somewhat different. After a steep increase observed in the beginning of the titanium ethoxide pulse, the ®lm mass continues to rise with a nearly constant rate. As the rate of the unsaturated deposition increased with the substrate temperature,

J. Aarik et al. / Thin Solid Films 370 (2000) 163±172

Fig. 1. Mass sensor signal as a function of time recorded (a) at substrate temperatures of 350 and 2158C during a single ALD cycle and (b) at substrate temperatures of 300 and 1508C during series of titanium ethoxide pulses (the upper curves) and a series of ALD cycles (the lowest curve).

the most probable reason of that is decomposition of titanium ethoxide. Indeed, as the O/Ti ratio is as high as four in a Ti(OCH2CH3)4 molecule, titanium dioxide can grow in the decomposition process without supplying any other oxygen

165

precursor. The unsaturated adsorption could also be caused by hydrolysis of surface intermediate species or gaseous titanium ethoxide by water vapor that may reside in very small amounts in the carrier gas. However, as will be discussed in Section 3.2, the latter process unlikely affects the growth in our experiments. Fig. 1a demonstrates that a small but clearly observable mass decrease occurs during the purge time used after the titanium ethoxide pulse. Possible reasons for this effect are desorption of surface intermediate species formed during adsorption of titanium ethoxide and/or desorption of decomposition products of these surface intermediates. Even faster decrease in the mass sensor signal appears when the water pulse is switched on. Consequently a signi®cant amount of reaction products desorbs from the surface in this reaction step. Thus the surface intermediate species formed during the titanium ethoxide adsorption contain ligands that are removed and/or decomposed by water pulse. As can be seen in Fig. 1a, during the water pulse the mass stabilizes on the level, which is higher than that before starting the ALD cycle. After switching off the H2O pulse the mass sensor shows no remarkable changes. This fact con®rms that the reaction products able to desorb from the surface are completely removed, no signi®cant amount of excess water adsorbs on the ®lm surface during the H2O pulse, and the solid reaction product causing the mass increment Dm0 is stable. The behavior of the ®lm mass recorded during a series of ALD cycles is depicted in Fig. 1b (the lowest curve). For comparison, the effect of successive titanium ethoxide pulses is also shown. One can see in the ®gure that the adsorption capability of substrate surface rapidly decreases if the titanium ethoxide pulses do not alternate with the H2O pulses. At 1508C, for instance, no adsorption can be recorded during the second titanium ethoxide pulse, already. At the same time the surface exposed to H2O readily adsorbs titanium ethoxide. No qualitative changes in the adsorption of titanium ethoxide occurred even if the purge time used after a H2O pulse was prolonged up to several hours. Therefore the steep mass increase that appears after switching on the titanium ethoxide pulse, is a result of adsorption rather than gas phase reactions which, in principle, may contribute to the growth when the titanium ethoxide and water pulses overlap with each other. Obviously the effect of overlapping is sensitive to the length of the purge times. It was established that the increase in Dm0 was below 5 and 15% when the purge times following the titanium ethoxide and water pulses, respectively, were reduced from 10 to 2 s. By contrast, Dm0 increased by 40±60% when the purge was decreased from 2 to 0.1 s whereas the most signi®cant increase in Dm0 appeared at the purge times below 0.5 s. Therefore overlapping of the precursor pulses obviously contributes to the ®lm growth at short purge times but its effect is insigni®cant at the purge times of about 2 s and longer. The data illustrating the dependence of Dm0 on different

166

J. Aarik et al. / Thin Solid Films 370 (2000) 163±172

Fig. 2. Mass changes (a) Dm0 and (b) Dm2 as functions of substrate temperature. Titanium ethoxide source temperature is 718C and H2O pressure is 6 Pa.

process parameters are presented in Fig. 2a. The ®gure shows that the variations of Dm0 are within ^20% when the substrate temperature rises from 100 to 2508C. Also, the effect of titanium ethoxide pulse time and following purge time is rather weak at these temperatures. At temperatures 200±2508C, however, a noticeable increase of Dm0 with the substrate temperature starts. Also a clearly observable in¯uence of the titanium ethoxide pulse time on Dm0 appears at these and especially at higher temperatures. However, the in¯uence of the purge time is still smaller than it could be expected taking into account the mass decrease observed at 3508C during the purge time applied after titanium ethoxide pulse (Fig. 1a). Therefore it is reasonable to conclude that the mass decrease observed during the purge time is caused by desorption of the titanium precursor ligands and/or decomposition products of them, only, rather than by desorption of titanium-containing surface intermediate species. An interesting result is that neither the titanium ethoxide

pulse time nor the following purge time affect the mass decrease Dm2 (Fig. 2b) even at the substrate temperatures where the Dm0 values obviously depend on the titanium ethoxide pulse time. It was checked at 3508C that neither the H2O pulse time varied from 1 to 10 s in¯uenced the Dm2 value. Consequently, at this temperature, the surface abundance of titanium precursor ligands, which can be removed during water pulse and/or preceding purge time, completely saturates with increasing titanium ethoxide pulse time although the amount of adsorbed titanium does not. Fig. 2b shows that Dm2 as a function of the substrate temperature has a maximum at about 2758C. Therefore at this temperature, the largest amount of ligands is removed during the H2O pulse. The main reason of the decrease in this amount at higher temperatures is that due to thermal decomposition, more ligands are released or decomposed during titanium ethoxide adsorption, already. At temperatures below 225±2508C the decrease of Dm2 can be explained by exchange reactions which probably occur on the surface during titanium ethoxide adsorption. It is generally known that the oxide surfaces exposed to water vapor are terminated with hydroxyl groups at suf®ciently low temperatures. Provided that adsorbing titanium ethoxide reacts with surface hydroxyl groups [13], some ligands can be removed in adsorption process. Thus, an increasing role of the exchange reaction between surface hydroxyl groups and adsorbing titanium ethoxide is one reason why the Dm2 value decreases at the substrate temperatures below 2508C. Another possible reason for smaller Dm2 values measured at low temperatures, is that the water pulse does not remove all precursor ligands from the ®lm surface. Indeed, Dm0 as a function of H2O pressure (Fig. 3a) and pulse time (Fig. 3b) does not saturate at substrate temperature of 1008C, although it saturates at 215 and 3508C. Similarly, no saturation of Dm2 was observed at T s ˆ 1008C. Instead, the Dm2/Dm0 ratio was almost invariant. These facts show that the reactions between surface intermediates and water vapor are not completed and small Dm2 values measured at temperatures close to 1008C may really be explained by incomplete removal of precursor ligands. Low reactivity of H2O towards the surface intermediate species is not surprising. For instance, in the tantalum oxide ALD process, the sticking coef®cient of water to the surface treated with tantalum ethoxide is about an order of magnitude smaller, than the sticking coef®cient of tantalum ethoxide to the surface treated with water [19]. Of course, low reactivity of H2O may, in principle, be compensated with higher pressure or longer pulse times. Indeed, signi®cant increase in the growth rate and obvious tendency towards saturation was observed at 1008C, as well, when the H2O pressure and pulse time exceeding 20 Pa and 5 s, respectively, were applied (Fig. 3). Unfortunately, longer purging is needed at higher H2O pressures in order to evacuate the excess of H2O from the reactor. Increase of the purge and pulse times, in turn, is usually not acceptable because of increasing deposition time. The data of QCM measurements demonstrate that at the substrate temperature of 2158C, Dm0 as a function of the

J. Aarik et al. / Thin Solid Films 370 (2000) 163±172

167

growth rate, depends on the substrate temperature. Higher source temperature is needed at T s ˆ 3508C where the saturation is achieved at higher Dm0 values than at T s ˆ 2158C (Fig. 4). This kind of behavior is related to the processes causing the delay between switching on the precursor pulse and the following mass sensor response (Fig. 1). The delay increases with the decrease in concentration of precursor molecules in the carrier gas [18]. The precursor concentration, in turn, decreases with decreasing source temperature. Thus, at a certain source temperature, the delay reaches the pulse length. Naturally, no saturation can be obtained at this and lower source temperatures, any more. The delay also increases with adsorption capacity of the surface [18] i.e. with the increase of the Dm0 values at which the saturation is obtained. For this reason it is not surprising that at higher saturation levels of Dm0, higher source temperatures are needed to keep the delay time shorter than the pulse time. 3.2. Properties of thin ®lms

Fig. 3. Dm0 as a function of (a) H2O pressure and (b) H2O pulse time recorded at substrate temperatures of 100, 215 and 3508C. Titanium ethoxide source temperature and pulse time are 788C and 2 s, respectively.

titanium ethoxide source temperature saturates when the latter reaches 65±708C (Fig. 4). However, the saturation is not complete when the substrate temperature is increased up to 3508C. This result is not surprising because, as mentioned above, a clearly observable unsaturated deposition occurs at this substrate temperature (Fig. 1a). Nevertheless, the variation of Dm0 is about 25%, only, when the titanium ethoxide source temperature rises from 65 to 858C (Fig. 4). An explanation for the weak dependence of Dm0 on the source temperature is that the unsaturated growth, obviously caused by decomposition of titanium ethoxide, is a surface-controlled process [13]. Indeed, if decomposition is slow compared with the adsorption of precursor molecules from the gas phase, then the surface becomes completely covered with precursor and in the ®rst approximation, the decomposition rate is determined by the substrate temperature rather than by the pressure of the precursor in the gas phase. An interesting feature shown in Fig. 4 is that the source temperature, at which the saturation effects appear in the

The thin ®lms for post-growth studies were grown with a set-up determined from the results of QCM measurements. The pulse and purge times were chosen equal to 2 s. The titanium ethoxide source temperature was set at 788C and the partial pressure of water vapor was kept at 20 Pa in the reactor during the water pulse. The set-up enabled us to grow ®lms with rather uniform thickness independently of the substrate temperature. Even at the temperature of 3508C, the thickness gradient did not exceed 10% on the 60-mm long substrates although no saturation of the titanium ethoxide adsorption with increasing pulse time was obtained at this temperature. This fact enables us to conclude that the unsaturated adsorption shown in Fig. 1a is due to the surface-controlled decomposition of titanium ethoxide rather than due to the gas phase reaction between titanium

Fig. 4. Dm0 as a function of Ti(OCH2CH3)4 source temperature recorded at substrate temperatures of 215 and 3508C. Pulse times and purge times are equal to 2 s.

168

J. Aarik et al. / Thin Solid Films 370 (2000) 163±172

ethoxide and water vapor residues. Indeed, the amounts of water that might reside in the carrier gas and/or desorb from the reactor walls are rather small. Therefore they can unlikely support constant pressure along the reactor and to cause uniform vapor-phase deposition on the whole substrate under the conditions where the titanium ethoxide concentration is much higher than that of the water vapor. The growth rate of the ®lms determined from thickness measurements, is 0.06 and 0.08 nm per cycle at substrate temperatures of 100 and 3508C, respectively. The increase of the growth rate at 3508C is in a good agreement with the dependence of Dm0 on the reactor temperature (Fig. 2a) measured with QCM. However, the growth rate of 300 nm thick ®lms grown at 2008C was 0.08 nm per cycle, i.e. about 30% higher than the value predicted from the growth rate values measured at 100 and 3508C, and from the temperature dependence of Dm0 (Fig. 2a). This difference is evidently connected with the surface roughness of the ®lms and will be discussed below. XRD and RHEED studies showed that the ®lms grown at substrate temperatures of 100±1508C were amorphous while those deposited at 1808C and higher substrate temperatures contained crystallites of the anatase structure (Fig. 5). According to RHEED patterns (Fig. 5a) the crystallites were randomly oriented in the ®lms grown at 1808C. Intense background indicated a rather high amount of the amorphous phase in these ®lms. Expectedly, the background intensity decreased with increasing growth temperature

Fig. 5. RHEED patterns of ®lms grown (a) by applying 1800 cycles at substrate temperature of 1808C and (b) by applying 3600 cycles at substrate temperature of 2758C.

and ®lm thickness. Moreover, the ®lms deposited at 2758C (Fig. 5b) and higher substrate temperatures showed a rather well developed preferential orientation. In these ®lms, the (110) plane of crystallites was preferentially parallel to the substrate surface. The sizes of crystallites depended on the deposition temperature, ®lm thickness and orientation of crystallites. In the ®lms of 200±300 nm thicknesses, the average sizes of crystallites ranged from 20 to 80 nm. A signi®cant amount of carbon (6±12 at.%) was recorded with AES and EPMA methods in the ®lms grown at 1008C. Also the O/Ti ratio was about 7% lower than that recorded for the ®lms grown at 1508C and higher substrate temperatures. These results indicated that the exchange reactions between water vapor and surface intermediate species were not complete at so low temperature. However, the carbon concentration rapidly decreased with increasing deposition temperature. According to AES data the carbon contamination did not exceed 0.5±0.7 at.% in the ®lms grown at T s ˆ 150±2008C while the concentration of 0.7 at.% was measured with the EPMA method for a ®lm deposited at 2008C. Optical measurements revealed that the refractive index of the ®lms grown at 1008C reached 2.3 and the extinction coef®cient was as low as 1 £ 10 23 at the wavelength of 580 nm. The ®lms grown at 3508C possessed the refractive index values of 2.4±2.5 and the extinction coef®cient about 8 £ 10 23. With decreasing ®lm thickness the value of refractive index somewhat increased and that of extinction coef®cient decreased. This kind of thickness dependence of optical constants indicated that the surface roughness probably increased with the ®lm thickness and in¯uenced the material parameters determined from the transmission spectra. The AFM measurements performed for 240±290 nm thick ®lms revealed that the root-mean-square value of the surface roughness was about 10 nm in the ®lms grown at 3508C (Fig. 6a). The surface roughness monotonously increased up to 25 nm when the growth temperature decreased to 2008C (Fig. 6b). With further decrease in the growth temperature, however, the roughness abruptly decreased and did not exceed 1 nm in the ®lms grown at 100±1508C (Fig. 6c). Therefore the surface roughness as a function of substrate temperature (Fig. 7) had a maximum at 2008C, that is very close to the lowest temperature, at which the crystallization appears. A very similar result has been observed earlier in case of TiO2 thin ®lms grown from TiCl4 and H2O [5]. High surface roughness explains why the growth rate determined for the deposition temperature of 2008C from the thickness of relatively thick ®lms is greater than the growth rate estimated from the QCM data. On one hand the ®lm density obviously decreases with increasing surface roughness. For this reason the same mass increase results in higher thickness increase. On the other hand the effective surface area also increases with surface roughness. Thus, rougher surface adsorbs greater amounts of precursors. As this effect increases with ®lm thickness, it has insigni®cant effect on the QCM measurements performed

J. Aarik et al. / Thin Solid Films 370 (2000) 163±172

169

at the ®lm thickness of 2±20 nm. However, surface roughening accelerates the growth of much thicker ®lms.

Fig. 7. Surface roughness as a function of substrate temperature used for thin ®lm growth. Film thicknesses range from 240 to 290 nm.

3.3. Growth mechanism The data of composition studies demonstrate that TiO2 thin ®lms with rather low concentration of impurities can be grown from titanium ethoxide and water at substrate temperatures 150±3508C. This fact together with the data of QCM measurements enables us to describe the growth mechanism in more detail. For instance, using the molar mass of TiO2 and the mass ratio Dm0/Dm1 (Fig. 8) where Dm1 ˆ Dm0 1 Dm2 (Fig. 1a), one can estimate the molar mass of surface intermediate species utilized in formation of each TiO2 unit during the water pulse. Furthermore, as will be demonstrated below, the Dm0/Dm1 ratio and its dependence on the process parameters permits to propose a model of surface reactions describing the ALD-type growth of TiO2 from titanium ethoxide and water at substrate temperatures where the self-limited growth is dominating. Some possible mechanisms of surface exchange reactions in ALD growth of TiO2 from titanium ethoxide and H2O have been discussed in an earlier paper [13]. Provided that the surface is at least partially hydroxyl-terminated, the

Fig. 6. AFM images of thin ®lms grown at substrate temperatures of (a) 350, (b) 200 and (c) 1508C.

Fig. 8. Dm0/Dm1 as a function of substrate temperature. Titanium ethoxide source temperature is 718C. Partial pressure of H2O is 6 Pa.

170

J. Aarik et al. / Thin Solid Films 370 (2000) 163±172

chemisorption of titanium ethoxide can be described as n…2OH† …s† 1 Ti…OCH2 CH3 †4 …g† ! …2O2†n Ti…OCH2 CH3 †42n …s† 1 nCH3 CH2 OH …g† …1† where (g) and (s) denote gas phase and surface species, respectively and n is the average number of surface hydroxyls reacting with a titanium ethoxide molecule in the adsorption process. In the case of hydroxylated surface 0 , n # 4. The following H2O pulse should result in the hydrolysis of the surface intermediates and recover the concentration of the hydroxyl groups on the ®lm surface. The reaction can be written as …2O2†n Ti…OCH2 CH3 †42n …s† 1 …4 2 n†H2 O …g† ! …2O2†n Ti…OH†42n …s† 1 …4 2 n†CH3 CH2 OH …g†

…2†

when all the ligands are replaced by hydroxyl groups. One can easily see from Eqs. (1) and (2) that the surface concentration of hydroxyl groups should continuously increase or decrease when the number of them consumed during the titanium ethoxide pulse (Eq. (1)) is not equal to those formed during H2O pulse (Eq. (2)). Therefore in case of a stable ALD growth when the growth rate does not depend on the number of cycles applied, and no hydroxyl groups remain in the ®lms, the value of n must be equal to two. Using the molar masses of substances, one can ®nd from Eqs. (1) and (2) that Dm0 =Dm1 ˆ 0:59 at n ˆ 2. This is signi®cantly higher than the experimental values (Fig. 8) measured for the substrate temperatures of 100±2508C, at which no decomposition of titanium ethoxide appears. The Dm0/Dm1 value can be lower than that calculated from Eqs. (1) and (2), if some titanium-containing surface intermediates formed during the titanium ethoxide pulse desorb before the water exposure. However, as mentioned above, no dependence of the growth rate on the purge length was observed. Therefore the contribution of this kind of desorption is insigni®cant. Another explanation for lower Dm0/Dm1 values is that less than two hydroxyls are involved in the chemisorption reaction described by Eq. (1). It is possible that titanium ethoxide is adsorbed on a completely dehydroxylated oxide surface as well [13] Ti…OCH2 CH3 †4 …g† ! Ti…OCH2 CH3 †4 …s†

…3†

In this case the following hydrolysis reaction should again result in a completely hydroxyl-free surface [13] Ti…OCH2 CH3 †4 …s† 1 2H2 O …g† ! TiO2 …s† 1 4CH3 CH2 OH …g†

…4†

One can ®nd that the Dm0/Dm1 value calculated from Eqs. (3) and (4) is equal to 0.35. This value is rather close to but somewhat lower, already, than the lowest experimental points in Fig. 8. Therefore an exchange reaction should still take place in the real growth process during titanium

ethoxide adsorption but less than two ligands, in average, should be removed from every titanium ethoxide molecule adsorbed. This kind of reaction can be described by Eq. (1) where 0 , n , 2. As mentioned above, a constant growth rate and deposition of the stoichiometric oxide can be achieved when the amount of hydroxyl groups consumed during the titanium ethoxide adsorption is equal to that formed during the water pulse. In case of n , 2 this means that only a fraction of the -OCH2CH3 ligands coordinated to titanium in the surface intermediate layer (see Eq. (1)) are replaced by hydroxyl groups in the hydrolysis process. The rest of the ligands are substituted by oxygen bridging between adjacent titanium atoms. As a result, the ®lm surface becomes partially hydroxyl-terminated at the beginning of the next ALD cycle. The corresponding exchange reaction can be written as …2O2†n Ti…OCH2 CH3 †42n …s† 1 2H2 O …g† ! …2O2†n TiO22n …OH†n …s† 1 …4 2 n†CH3 CH2 OH …g† …5† where 0 # n # 2. The Dm0/Dm1 ratio calculated from Eqs. (1) and (5) depends on n and can be expressed as Dm0 =Dm1 ˆ 79:9=…227:9 2 46n†

…6†

Using the experimental data presented in Fig. 8 and Eq. (6) one can estimate the average value of n for each substrate temperature at which the reactions are self-limited. Calculations show that n is equal to 0.6 at the substrate temperatures of 225±2508C, at the highest temperatures where the decomposition of titanium ethoxide is negligible. At temperatures below 2258C, the decrease in Ts is accompanied with the increase in Dm0/Dm1 (Fig. 8). As a result, the value of n calculated from Eq. (6), increases up to 1.2 and 1.5 at 200 and 1508C, respectively. Unfortunately, at temperatures below 1508C, this kind of estimation of the n value is not reliable any more because the exchange reactions are not completed during the water pulses and the ®lms contain a signi®cant amount of carbon. Moreover, the decrease of the O/Ti ratio accompanying the increase of carbon contamination in the ®lms grown at 1008C can not be explained by the deposition mechanisms discussed. Thus, the reactions are obviously more complex at so low temperature. Also, the model proposed does not apply at the substrate temperatures where titanium ethoxide decomposes and unsaturated deposition appears. The observation that only 0.6±1.5 ligands, in average, are released from each titanium ethoxide molecule during its adsorption at substrate temperatures 150±2508C is in a good agreement with in situ mass spectrometry studies of the processes where metal ethoxides (Ti(OCH2CH3)4, Ta(OCH2CH3)4 and Nb(OCH2CH3)4) and water have been used as the precursors [20]. The data of these studies indicate that CH3CH2OH is almost exclusively released during the water pulse. The small amount of ligands released during adsorption of metal precursor is one of the main reasons limiting the

J. Aarik et al. / Thin Solid Films 370 (2000) 163±172

growth rate. Indeed, if in average, 2.5±3.4 relatively large OCH2CH3 ligands are adsorbed together with each titanium atom, then the concentration of titanium in the surface intermediate layer is determined by the size of the surface intermediate species rather than by the density of titanium sites on the surface of TiO2. Incomplete exchange of ligands in the low-temperature hydrolysis process is another reason for low growth rate. The ligands not removed during the water pulse stay on the ®lm surface and diminish the number of adsorption sites for titanium ethoxide in the next ALD cycle. On the other hand, incomplete removal of the ligands from the ®lm surface enhances the probability for the incorporation of the ligand constituents into the ®lm. Therefore carbon contamination of the ®lms grown at the substrate temperature of 1008C con®rms that at this temperature, the water doses used are really unable to remove all ligands from surface intermediate layer. Despite possible limitations, the growth rate obtained in the present work is remarkably higher than that observed for a similar ALD process earlier [13]. At the substrate temperatures 150±3508C, the growth rate well compares to that achieved for the TiCl4/H2O and TiCl4/H2O2 ALD processes in the same temperature range [3,5,9,10]. Furthermore, according to our preliminary studies, the structure and optical properties of the ®lms investigated in this work compare to those of the thin ®lms deposited by ALD from TiCl4 and H2O at similar substrate temperatures [5,8]. Finally it should be noted that the changes in the thin ®lm structure appearing at substrate temperatures of about 1808C evidently in¯uence the surface reactions as well. In this connection, one should take into account that the real-time QCM studies have been performed at a rather small thickness of the ®lm deposited onto the mass sensor. Typically the thickness was below 20 nm. Furthermore, the deposition conditions were signi®cantly varied in the growth process. It is clear that no highly developed crystal structure can be obtained under these conditions. For this reason the mechanisms discussed above adequately describe the growth of amorphous TiO2 and/or the initial stage of nonepitaxial growth of anatase. The increase of thickness usually results in changes in surface roughness and crystallite orientation of polycrystalline ®lms [5] and, in this way, may affect adsorption processes. Therefore the reaction mechanisms may somewhat change in the growth process. Nonetheless, a rather good agreement between the results of real-time and post-growth measurements is obtained. According to rough estimations the variation in growth rate caused by crystallization and surface roughening do not exceed 30% even if the ®lm thickness reaches 300 nm.

could be reproducibly deposited from these precursors at the substrate temperatures ranging from 100 to 3508C. The adsorption of titanium ethoxide was self-limited at the substrate temperatures up to 225±2508C. Although the decomposition of titanium ethoxide signi®cantly contributed to the ®lm growth at temperatures above 2508C, the effect of titanium ethoxide source temperature on the growth rate was weak and ®lms of uniform thickness were obtained even under these conditions. QCM studies revealed that at the substrate temperatures of 225±2508C, low contribution of exchange reactions during titanium ethoxide adsorption was a reason for limited growth rate. According to our estimations less than one ligand, in average, was liberated from each titanium ethoxide molecule adsorbed at these temperatures. Correspondingly, each surface intermediate complex consumed for formation of one TiO2 unit during the following water pulse, contained more than three ligands. For this reason each of surface intermediate species was able to cover several titanium sites on the oxide surface and the growth rate was evidently limited by the steric hindrance. At substrate temperatures below 1508C, the growth rate was affected by the insuf®cient reactivity of H2O towards the surface intermediate species formed during the titanium ethoxide pulse. No complete saturation of the growth rate with increasing water dose was achieved and the ®lms contained carbon residues. Nevertheless, the QCM and optical thickness measurements showed that at suf®cient H2O doses, the growth rate was about 0.06 nm per cycle at the substrate temperatures of 1008C and did not change remarkably when the temperature was raised to 2508C. The ®lms grown at 100±1508C were amorphous while those deposited at 1808C and higher substrate temperatures were polycrystalline. Anatase was the only crystalline phase observed in the ®lms with X-ray diffraction. The surface roughness was highest in case of ®lms grown at the substrate temperature of 2008C, i.e. very close to the minimum temperature at which the crystallization occurred. The refractive index of amorphous ®lms grown at 1008C was 2.3 while the extinction coef®cient equaled to 1 £ 10 23 at the wavelength of 580 nm. The refractive index of the polycrystalline ®lms grown at 3508C reached 2.5 and the extinction coef®cient was about 8 £ 10 23. Acknowledgements The authors are grateful to Alma-Asta Kiisler and Jelena Asari for technical assistance, and to Hugo MaÈndar for XRD measurements. The work was supported by Finnish National Technology Agency (TEKES) and Estonian Science Foundation (Research Grants No. 1878 and 3871).

4. Conclusions Real-time studies of the TiO2 growth from titanium ethoxide and water demonstrated that titanium dioxide thin ®lms

171

References [1] T. Suntola, Mater. Sci. Rep. 4 (1989) 265.

172

J. Aarik et al. / Thin Solid Films 370 (2000) 163±172

[2] V.B. Aleskovskii, V.E. Drozd, Acta Polytech, Scand. Chem. Tech. Ser. 195 (1990) 155. [3] M. Ritala, M. LeskelaÈ, E. NykaÈnen, P. Soininen, L. NiinistoÈ, Thin Solid Films 225 (1993) 288. [4] M. Ritala, M. LeskelaÈ, L.-S. Johansson, L. NiinistoÈ, Thin Solid Films 228 (1993) 32. [5] J. Aarik, A. Aidla, T. Uustare, V. Sammelselg, J. Cryst. Growth 148 (1995) 268. [6] J. Aarik, A. Aidla, V. Sammelselg, H. Siimon, T. Uustare, J. Cryst. Growth (1996) 169. [7] N.V. Dolgushev, A.A. Malkov, A.A. Malygin, et al., Thin Solid Films 293 (1997) 91. [8] J. Aarik, A. Aidla, A.-A. Kiisler, T. Uustare, V. Sammelselg, Thin Solid Films 305 (1997) 270. [9] V. Sammelselg, A. Rosental, A. Tarre, et al., Appl. Surf. Sci. 134 (1998) 78. [10] H. Kumagai, M. Matsumoto, K. Toyoda, M. Obara, M. Suzuki, Thin Solid Films 263 (1995) 47.

[11] H. DoÈring, K. Hashimoto, A. Fujishima, Ber. Bunsenges. Phys. Chem. 96 (1992) 620. [12] M. Ritala, M. LeskelaÈ, L. NiinistoÈ, P. Haussalo, Chem. Mater. 5 (1993) 1174. [13] M. Ritala, M. LeskelaÈ, E. Rauhala, Chem. Mater. 6 (1994) 556. [14] H. Siimon, J. Aarik, J. Phys. D: Appl. Phys. 30 (1997) 1725. [15] J. Aarik, A. Aidla, A.-A. Kiisler, T. Uustare, V. Sammelselg, Thin Solid Films 340 (1999) 110. [16] V. Sammelselg, J. Aarik, A. Aidla, et al., J. Anal. At. Spectrom. 11 (1999) 523. [17] R. Swanepoel, J. Phys. E: Sci. Instrum. 16 (1983) 1214. [18] J. Aarik, H. Siimon, Appl. Surf. Sci. 81 (1994) 281. [19] K. Kukli, J. Aarik, A. Aidla, H. Siimon, M. Ritala, M. LeskelaÈ, Appl. Surf. Sci. 112 (1997) 236. [20] M. Ritala, M. Juppo, K. Kukli, A. Rahtu, M. LeskelaÈ, J. Phys. IV 9 (1999) Pr8±1021.

Suggest Documents