Article pubs.acs.org/cm
Growth and Crystallization of TiO2 Thin Films by Atomic Layer Deposition Using a Novel Amido Guanidinate Titanium Source and Tetrakis-dimethylamido-titanium Marcel Reiners,*,† Ke Xu,‡ Nabeel Aslam,† Anjana Devi,‡ Rainer Waser,†,§ and Susanne Hoffmann-Eifert† †
Forschungszentrum Jülich, Peter‐Grünberg‐Institut und Jülich Aachen Research Alliance on Fundamentals for Future Information Technology (JARA‐FIT), 52425 Jülich, Germany ‡ Inorganic Materials Chemistry, Ruhr University Bochum, 44780 Bochum, Germany § Institut für Werkstoffe der Elektrotechnik 2, RWTH Aachen University, 52074 Aachen, Germany S Supporting Information *
ABSTRACT: We studied the growth of TiO2 by liquid injection atomic layer deposition (ALD) utilizing two different amide-based titanium sources, tetrakis-dimethylamidotitanium [(NMe2) 4 -Ti, TDMAT] and its recently developed derivative, tris(dimethylamido)-mono-(N,N′-diisopropyl-dimethyl-amido-guanidinato)-titanium {[(NiPr)2NMe2]Ti(NMe2)3, TiA3G1}, with water vapor as counterreactant. A clear saturation of growth with an increasing precursor supply was found for TDMAT between 150 and 300 °C and for TiA3G1 between 150 and 330 °C. Representative growth per cycle (GPC) values at 250 °C were 0.041 and 0.044 nm/cycle, respectively. Compared to that of TDMAT, ALD of TiA3G1 exhibited a significantly higher stability in the GPC values up to 300 °C coinciding with an improved temperature stability of the precursor. Both processes showed a minimum of the growth rate as a function of temperature. In all cases, the residual carbon and nitrogen contents of the TiO2 films were 34 nm exhibit additional rutile crystallites that were identified by the R(110) reflection at 27.8°.51 The growth of rutile TiO2 at only 300 °C on a nonmatching substrate is remarkable. In general, the growth of rutile TiO2 was shown to need initiation by the presence of the rutile structure of the layer beneath it.52 Jogi et al. reported on the growth of rutile TiO2 from TiEt5 and H2O on Si(100), but the estimated growth temperature was 350 °C.50 A possible explanation of why we already achieved the rutile phase of TiO2 at the lower temperature of 300 °C is given in the next section. The GIXRD pattern of TiO2 films from TiA3G1 and H2O is additionally shown from 250 to 330 °C in Figure 5a. For TiO2 films deposited at 250 °C, only the diffraction peak of anatase phase A(101) was observed. The rather high intensity of the reflection indicates either a large number of crystallites or a significant increase in the size of the crystallite. For higher growth temperatures, the A(101) peak already appears at lower film thickness, while the maximal intensity for a thicker film (∼50 nm) remains below the A(101) intensity of the 35 nm thick film at 250 °C. The change in crystallization behavior is also obvious from the appearance of additional peaks in the GIXRD spectra of TiA3G1-derived TiO2 films, which were grown at temperatures of ≥300 °C. While brookite phase B(121) at 30.8° was already observed for the TDMAT process at 250 °C, its development was retarded to 300 °C when the TiA3G1 ALD process was used. The brookite phase vanished again at higher deposition temperatures. A similar type of retardation was observed for the rutile phase, where the first reflections were observed at 315 °C. Here the peak height of the A(101) reflection decreased with an increase in temperature in favor of the R(110) peak height corresponding to a higher fraction of the rutile crystallites with respect to the total amount of all kinds of crystallites. Additionally, AFM measurements were conducted to reveal the influence of the growth parameters on the surface morphology with the focus on crystallization behavior. In Figure 6, the AFM images of TiO2 films from TDMAT grown at 250 and 300 °C are arranged. Along each row, films of increasing thickness are shown grown with 250, 600, and 800 ALD cycles. Panels a−c of Figure 6 show the evolution of the
Figure 6. AFM images of TiO2 on SiO2 from TDMAT and water. Panels a−c show the increase in the density of surface objects on TiO2 films grown at 250 °C as a function of ALD cycle number. Panels d−f show the corresponding AFM images for 300 °C revealing an increased number of smaller objects.
surface morphology for films deposited at 250 °C. In agreement with the observation that quasi-amorphous thin films were grown at 250 °C, the films exhibit a smooth morphology. At greater thicknesses, or higher numbers of ALD cycles, the density of grainlike structures increases and is accompanied by very small dotlike objects. The ALD cycle series for 300 °C is shown in Figure 6d−f. Here the surface morphology changed in such a way that only very small surface objects appeared throughout the whole ALD cycle series. This confirms the findings from GIXRD analysis that for an increased deposition temperature the crystallization already starts at earlier stages of growth. The AFM images of TiO2 films from TiA3G1 and water are depicted in Figure 7a−i in the same manner, assigning each ALD cycle series to one line for deposition temperatures from 250 to 330 °C. For a growth temperature of 250 °C, TiO2 films from TiA3G1 that exhibited large anatase-type crystallites showed the evolution of more regularly shaped objects in contrast to TiO2 grown by TDMAT. These pyramid-like objects are assigned to anatase crystallites.53,54 Clearer evidence of this assignment was taken from a detailed AFM and SEM analysis of TiO2 thin films grown by TTIP and water as shown in Figure 9a for comparison. The same ALD parameters were used as in our previous study, which reveals only anatase-grown TiO2 thin films.13 Generally, an increase of the deposition temperature leads to a refinement of the crystallite size as depicted in Figure 6d−f and 7d−i. The crystallite size, the difference in crystallinity, and the amount of crystallites lead to differences in root-mean-square (rms) roughness. In Figure 8, the rms roughness of TiO2 films grown by TDMAT and TiA3G1 is shown as a function of thickness and ALD cycle number, respectively, and temperature. Quasi-amorphous TiO2 films grown at 150−200 °C showed a constant rms roughness below 0.5 nm, irrespective of the films’ thickness. For a growth temperature of 250 °C where the density of crystallites increased with thickness during both processes, the quantitative analysis of the rms roughness shown in Figure 8 revealed a stronger increase in roughness for TiO2 films grown by TDMAT than for TiA3G1. This indicates that the pyramidF
dx.doi.org/10.1021/cm303703r | Chem. Mater. XXXX, XXX, XXX−XXX
Chemistry of Materials
Article
nearly linear dependence of the rms roughness on thin film thickness as seen for both processes in Figure 8. Crystallization Behavior. Comparing the results of the GIXRD and AFM studies for the two temperature−thickness series of ALD TiO2 films from TDMAT and TiA3G1, we discovered a clear difference in crystallization behavior with respect to the ALD process. This becomes obvious when the development of the intensity of the anatase A(101) reflection with an increase in temperature is compared for the two different processes and when, in addition, the retardation of the development of the brookite and the rutile phase was taken into account. At first sight, it seems to be controversial that the anatase A(101) reflection is more dominant in thin films deposited from TiA3G1 than in films grown with TDMAT, but this apparent contradiction can be understood if crystallization kinetics is considered. The time dependence of a nucleation process is generally described by the Johnson−Mehl−Avrami− Kolmogorov equation.55,56 This equation states that the fraction of the material that undergoes a phase transition is proportional to the time under isothermal conditions. In our case, this means that the time of an ALD cycle might have a significant effect on the crystallization behavior of the grown thin films. In detail, for our experiments, we had a difference in the time of one cycle of ∼2.25 s between the shorter TDMAT (19.5 s) and the longer TiA3G1 (21.75 s) cycles. In the case of 800 cycles, this summed to a difference in total process time of 30 min so that the TiA3G1 process took 30 min longer than the TDMAT process. Therefore, the difference in the intensity of the A(101) reflection for films deposited from the TiA3G1 and TDMAT processes becomes reasonable because processing at lower temperatures for longer times should favor the growth of the anatase phase. Additionally, also other crystallization-related effects such as the retardation of the development of the brookite phase, and similarly the development of the rutile phase, could be understood if the effect of process time on the crystallization behavior of the films is considered. The analysis of the temperature−thickness−time dependence of the crystallization of ALD-grown TiO2 films clearly shows that at longer time scales (i.e., for TiA3G1) the anatase phase is preferentially developed. This means that either more amorphous material is taken from the surroundings of the crystallites or crystallites of different phases may become swallowed up during the lateral growth of the anatase crystallites. Besides isothermal crystallization, the rate constant of crystallite growth plays an important role. The rate constant is proportional to the activation energy for crystallization, which is different for distinct crystalline phases of the same material. From this, the preferential growth of the rutile and brookite phases at higher temperatures at given process times can be understood. Therefore, the control of the ALD process time may be another important key parameter for controlling crystallization into the rutile phase on surfaces that do not already have the rutile structure to initiate quasi-heteroepitaxial growth. The scheme of ALD TiO2 film growth can also be extended to our previous study13 where the process time was as long as 38 s. These process conditions favored only the growth of the anatase phase, and a cycle-dependent transition of the GPC occurred. The transition in the growth rate was correlated to the onset of the crystallization of anatase-type grains, and the change in GPC behavior was attributed to an increase in hydroxylic group density on the grown anatase crystallites.13 This scenario is schematically depicted in Figure 9b. As a consequence of this growth scenario, we predict that any
Figure 7. AFM images of TiO2 grown by TiA3G1 and water at various ALD cycles from left to right and deposition temperatures from top to bottom. SiO2|Si was used as the substrate material. For 250 °C in panels a−c, the development of rectangular grains was observed with an increasing cycle number. From panel d to g, the surface morphology was refined to small grains with an increasing temperature.
Figure 8. rms roughness vs the number of ALD cycles for TiO2 films grown by TDMAT and water (bottom) and for films grown from TiA3G1 and water (top).
shaped anatase crystallites for the TiA3G1 process contribute less to the films’ rms roughness than the mixture of the brookite and anatase phase crystallites that were dominant for thin films grown by TDMAT at 250 °C. At ≥300 °C, at which the TiO2 films showed initial growth of crystallites, and assuming the density of crystallites to be constant during further stacking of crystalline material, the crystallites grew in size, resulting in a G
dx.doi.org/10.1021/cm303703r | Chem. Mater. XXXX, XXX, XXX−XXX
Chemistry of Materials
Article
Figure 10. Ti mass layer density vs the number of cycles of TiO2 from TDMAT and water at ALD cycle times of 19.5 s (circles) and 37.5 s (squares) at 260 °C. The green dotted line corresponds to the linear regression of the short time process. Red and blue lines are the fits for low and high cycle numbers, respectively, for the long time process. The inset shows the GIXRD pattern of TiO2 grown by the short and long process at different cycle numbers.
The discussion of the crystallization kinetics demonstrates the impact of time for crystallization on the different phases or phase mixtures of TiO2. In addition, with respect to crystallization time, care must be taken to remove the wafers from the reactor system immediately or to keep the remaining time in the reactor constant. Otherwise, a precise control of films’ thickness and microstructure cannot be ensured. To complete the model of crystallization, other effects have to be taken into account as well. In recent publications describing ALD TiO2 grown on different substrate materials such as Al2O3,54 Ti, and Si,53 only the anatase phase was detected at 250 °C for increased thicknesses and at ALD cycle times of