Fig. 3.3(a) compares the XRR curves of Al2O3/Ge samples deposited at 50 and 300 °C.
The four-layer model of the Al2O3/GexAl1–xO/GeO2/Ge gate stack fit the measured curve nearly perfectly; note that the presence of a GexAl1–xO interlayer—arising from intermixing of Al2O3 and GeO2—had to be considered into the model for accomplishing an excellent fit to the curve. The oscillation frequency is determined primarily by the thickness of the overlying Al2O3 film, and the oscillation amplitude is very sensitive to the variation in roughness between the top surface and the deposited interface. The 300 °C-Al2O3 on Ge substrate exhibited a slower oscillation frequency with a higher amplitude variation, reflecting its thinner Al2O3 film and rougher Ge interface, with respect to those of the substrate prepared at 50 °C. Note that the top surface roughness (<3 Å) was similar in all samples. We also observed that the peak positions of the maxima shifted with the incident angle in these two
samples, a typical effect of the Al2O3 density changing with the value Tdep. Indeed, the film density increased from ca. 3.3 g cm–3 to ca. 3.7 g cm–3 [Fig. 3.3(b)]. Compared with the high density (3.97 g cm–3) of crystalline α-Al2O3 [26], the density of each as-deposited Al2O3 film was indeed closer to those reported for γ-Al2O3 or amorphous alumina (3.5–3.7 g cm–3) [26], [27]. The lower densities of the Al2O3 films observed upon decreasing the value of Tdep arose mainly from the thermally activated reaction kinetics dominating at lower temperatures—especially because we used TMA and H2O in the ALD process, which caused an increase in the levels of H, OH, and C impurities in films. Using forward recoil spectrometry (FReS), the hydrogen concentration in ALD-Al2O3 has been observed to increase upon decreasing the value of Tdep from 175 to 35 °C [3]. SIMS analyses (Fig. 3.4) provided further evidence for a relatively large carbon concentration within the main Al2O3
prepared at temperatures below Tmax, i.e., <200 °C. When excess O atoms or the C/H contaminants were present within the Al2O3—probably in the form of aluminum hydroxide [Al(OH)3] and aluminum oxy-hydroxide, the densities of which are 2.42 and 3.44 g cm–3, respectively [28]—the film density was lowered accordingly.
Figure 3.3(c) depicts the respective thicknesses of main Al2O3, mixed Al2O3–GeO2, and interfacial GeO2 layers after 60 deposition cycles as a function of Tdep; at first, the overall XRR thicknesses were close to the TEM observations. Interestingly, as the value of Tdep
increased from 50 to 300 °C, the thickness of the underlying GeO2 layer decreased gradually from ca. 5 Å to below 1 Å, whereas that of the intermediate GexAl1–xO layer increased dramatically from ca. 1 Å to ca. 14 Å, accompanied by increased roughness between the GeO2 layer and Ge substrate. These structural degradations became severe at temperatures above 200 °C, implying that, during ALD deposition, the poorly oxidized GeOx
species at the surface were less stable and probably diffused into the top Al2O3 at higher values of Tdep, leading to more severe dielectric intermixing and degraded interfacial roughness. Seo et al. speculated that similar behavior occurred during the growth of HfO thin
films on Ge substrates using molecular beam epitaxy (MBE) [29]; they observed that the surface GeOx species possibly dissolve in HfO2 at a growth temperature of 360 °C.
Next, we examined the relevance of the Al 2p and O 1s photoemission spectra, in particular their peak energy spacing; Fig. 3.5(a) provides an example for the as-deposited Al2O3/Ge. The value of the energy spacing was ca. 456.95 eV, close to the value of 457 eV for sapphire, but we did not observe a correlation between the energy spacing and the deposition temperature because the chemical shift was within 1 eV. Therefore, we further investigated the stoichiometry of the temperature-dependent grown Al2O3 before and after rapid thermal processing [Fig. 3.5(b)]. Two noteworthy features are that (i) the O/Al composition ratio gradually achieved the ideal value of 1.5 upon increasing the value of Tdep to 200 °C, but it increased to 1.65 at 300 °C, and (ii) a more-stoichiometric film was formed after subsequent RTA. We can explain the variation of the O/Al composition ratio reasonably in terms of the controlled growth mechanism. As we mentioned earlier, raising the value of Tdep led to redundant oxygen-based radicals being expelled from the deposited films, which in turn suppressed the oxygen content and, thus, improved the Al2O3 stoichiometry. Above Tmax (ca.
185 °C), the deposition process relies strongly on the presence of AlOH* or Al(CH3)* reactive sites because of the lower adsorption rate. In other words, a relatively higher H2O concentration exists close to the surface at higher Tdep, presumably leading to the oxygen-excessive Al2O3 film observed when Tdep was 300 °C. Similar tendency was also characterized in the electrical permittivity that are ca. 5.1, 6.2, 7.9, and 5.8 for as-deposited Al2O3 films grown at Tdep of 50, 100, 200, and 300 °C, respectively; a relatively higher value was obtained at 200 °C. Furthermore, we suggest that high-temperature annealing provides additional thermal energy, acting as an external driving force, to return these as-deposited Al2O3 samples to their stable, stoichiometric phase.
On the other hand, GeOx out-diffusion after thermal annealing is a critical issue in the preparation of high-k/Ge structures. Surface-sensitive Ge 2p3/2 spectroscopy [Fig. 3.6(a)]
revealed that no Ge oxides diffused into the top of the Al2O3 dielectric after 400 °C RTA, but the bulk-sensitive Ge 3d spectrum (inset) revealed an increased GeOx/Ge intensity ratio, indicative of GeO2 growth near the lower interface. When we increased the RTA temperature to 600 °C, we detected a small peak corresponding to GeO2 in the Ge 2p3/2 spectrum along with a reduced GeOx/Ge ratio in the Ge 3d spectrum. We speculate that the balance between two competing processes—oxide growth and oxide desorption, which are strongly dependent on both the temperature and oxygen concentration—determines the amount of residual Ge oxide and its distribution. Because we used a N2 ambient as the feed gas, oxide desorption (rather than oxide growth) should dominate the thermal reaction mechanism. Together with the finding that the critical temperature of GeO desorption was in the range of 360–425 °C [30], [31], it is probable that GeO2 continued to grow at ca. 400 °C due to the presence of some residual oxygen in the N2 ambient, but it most likely desorbed at the higher temperature of 600 °C. In Fig. 3.6(b), the Al2O3/Ge deposited at a lower temperature of 50 °C exhibited results similar to that deposited at 100 °C after RTA; moreover, the higher-temperature Al2O3
samples appeared to display lower resistivity in GeOx incorporation. Note that the calculated value of the GeO2 atomic concentration in Al2O3—based on a standard sampling depth of ca.
30 (±4) Å for the Ge 2p3/2 spectrum—might be affected by different Al2O3 overlaying thicknesses in these samples. Here, the GeO2 concentration remained low (<1.0 at.%) in all cases. In view of the earlier XRR examination, the as-deposited Al2O3/Ge structures displayed an increased GexAl1–xO intermixing phenomenon, especially at temperatures up to 200 °C.
Therefore, we conclude that high-temperature annealing is likely to cause interfacial GeOx
volatilization on the bottom of the Al2O3 dielectric with a small degree of Ge incorporation. It was suspected that Al2O3/Ge capacitors undergoing RTA at 600 °C might suffer from resultant dielectric intermixing, as in the case of the high-Tdep systems (≥200 °C); indeed, both the interface quality and the leakage current characteristics deteriorated (data not shown). These experimental findings agree with those of previous studies suggesting that the volatilization of
gaseous GeO into high-k layers degrades the insulator properties [6], [32].