Capping
6.1 Introduction
As the previous report in chapter 5, uniform nano-nucleus CoSi2 with nano island size of 4 nm can be achieved by using oxide-mediated silicidation process with Ti-capping in a lower vacuum environment process. In addition, a thicker epitaxial CoSi2 film of 40-50 nm formed by the OME method with a Ti-capping layer has been reported by Kim et al. [1,2]. However, the detailed mechanism was not provided. For the same structure, Detavernier et al. [3,4] have ascribed the enhancement of cobalt silicide formation to the reduction of SiOx to CoxTiyOz by Ti. In this study, we find that Ti can absorb oxygen from the SiOx layer and decompose it, which is different to that of Detavernier et al. [3,4].
6.2 Experimental Details
P-type (8-12 ohm-cm) silicon substrates were chemically cleaned and dipped in a 60 °C HCl:H2O2:H2O=3:1:1 solution for 1 min to form a SiOx layer. Subsequently, a Co layer was deposited by magnetron sputtering followed by a Ti-capping layer deposition in the same chamber with vacuum breaking to change targets. The Co and Ti targets (99.95% purity) were pre-sputtered for 10 minutes after the base pressure of 3x10-6 torr was reached using argon (99.995% purity) as the sputtering gas. The thicknesses of the Ti/Co/SiOx stack were determined to be about 14/7/0.8 nm from a TEM cross sectional image. In addition, another multilayer sample of TiN/Co/Ti/Co/SiOx was also prepared using the same method except that the time and temperature to form the SiOx layer was increased to 3 minutes and 90 °C to increase the thickness. The thicknesses of the TiN/Co/Ti/Co/SiOx multiplayer were determined to be about 8/3/6/3/2 nm. Ex-situ annealing was carried out in a vacuum chamber of 10-4 torr. Upon annealing, all layers of the latter sample except the reactive products were stripped off by chemical etching, in order to examine the silicide layer in plan-view. TiN, unreacted Co and SiOx layers can be stripped off by NH4OH:H2O2: H2O=1:1:4 solution at 50 °C, aluminum etching solution (H3PO4 71 wt%, HNO3 2.5 wt%, CH2COOH 12.5 wt%, others H2O) at 75 °C and HF solution, respectively. The phase of the samples was then examined by transmission electron microscopy (TEM).
The atomic redistribution of the TiN/Co/Ti/Co/SiOx multiplayer was examined by Auger electron spectroscopy (AES) depth profiling.
6.3 Results and Discussion
Fig. 6-1 shows a bright-field TEM cross sectional image of the as-deposited Ti/Co/SiOx sample. This image clearly shows that the SiOx layer was 0.8 nm in thickness and sandwiched between Co and Si, whereas the thicknesses of the Co and Ti layers are about 7 nm and 14 nm, respectively. Fig. 6-2 (a) shows a bright-field TEM cross sectional image of the sample after annealing at 460 °C for 120 sec. From Fig. 6-2 (a), Co has diffused through the SiOx layer and a wavy cobalt silicide was formed, indicating that the Co diffusion may be only through some weak points in the SiOx layer, which agrees with the results of Detavernier et al. [3,4]. One can also easily find that the SiOx layer becomes more discontinuous than the as-deposited sample in Fig. 6-1, implying that some SiOx disappears. Fig. 6-2 (b) is a close up high resolution TEM image of Fig. 6-2 (a). Apparently, lattice fringes and dark contrast in the SiOx layer verify that part of the layer has been reacted and crystallized. If the annealing time was further prolonged, the entire SiOx layer is gone as shown in Fig.
6-3, which is a bright-field TEM cross sectional image of the sample after 460 °C 300 sec annealing. According to the study of Baten and Fedorovich [5,6], Co diffuses
through SiO2 without any chemical interaction with the SiO2 networks, but only occupies interstices of the very open SiO2 structure and migrates along the interstices as diffusion channels without affecting the regular lattices. This indicates that the SiOx layer would not react with Co and would not be destructed by the Co diffusion.
But why does the SiOx layer disappear? Detavernier et al. have reported [3,4] that Ti should diffuse through the grain boundaries of the Co layer to the SiOx layer, and reduce SiOx to CoxTiyOz, which then enhances the Co diffusion to form cobalt silicide.
However, in their study, the SiOx layer was still left after annealing because of a thicker SiOx layer, so this cannot explain our results either. It is also noted that another white line appears between the Ti and Co layers in Fig. 6-3, which should be a Ti-oxide layer resulting from absorbing oxygen from environment and the Co layer according to Detavernier et al. [3,4]. The residual oxygen in the Co layer is due to exposure to air prior to the Ti-capping layer deposition. Could the SiOx layer also be decomposed by Ti because Ti has the larger negative oxide formation enthalpy than Si and Co (see Table 6-1) [5,7]? In order to prove this assertion, another multilayered sample of TiN/Co/Ti/Co/SiOx/Si structure was prepared without vacuum break between layers to prevent oxygen contamination, and the thickness of the SiOx layer was increased to about 2 nm so as to easier monitor oxygen diffusion from the SiOx
layer.
Fig. 6-4 (a) is a bright-field TEM cross sectional image of this multilayered sample after 600 °C 240 sec annealing, significantly higher temperature than the previous one. First of all, no extra white line was found in the Ti and Co interface, which illustrates the residual oxygen in the Co layer has been eliminated by this deposition process. Fig. 6-4 (b) is a TEM bright field plan-view image with a diffraction pattern from the reactive silicide of the same sample as in Fig. 6-4(a) after all the other layers have been removed. The diffraction pattern shows that the silicide is CoSi2, in agreement with our previous results in that the CoSi2 is the first phase from the oxide-mediated cobalt silicidation, which also confirms that the dark contrast in Si adjacent to SiOx is cobalt silicide in Fig. 6-4(a). Fig. 6-4 (c) shows the AES depth profiles of the sample in Fig. 6-4 (a), where every layer is clearly distinct and oxygen from the SiOx layer was largely absorbed in the Ti layer after annealing. This directly supports our above supposition and is different from the mechanism proposed by Detavernier et al. [3,4]. This discovery that SiOx layer can be decomposed by Ti and even disappearing overthrows the impressions that the SiOx will keep remaining on the cobalt silicide layer after annealing in the OME process [8-11]. This can also explain why a CoSi2 layer from a Ti-capped multilayer produced by a single OME process can be thicker than ~11nm [12,13]. That is because the Ti capping layer thins the SiOx layer to increase the Co diffusion rate while the SiOx layer can reduce the Co
diffusion rate to form a CoSi2 layer from a non-Ti-capped multilayer. By comparing Fig. 6-3 with Fig. 6-4 (a) for both samples comprised of the Ti capping layer, the thicker SiOx layer in Fig. 6-4 (a) more effectively deters the Co diffusion rate resulting in a thinner cobalt silicide layer than Fig 6-3, although the sample in Fig. 6-4 (a) experiences annealing at a higher temperature.
The sample in Fig. 6-3 was annealed further at 600 °C for 300 sec and the result is shown in Fig. 6-5. Apparently, because the SiOx barrier layer is gone, Co reacts directly with the Si substrate as the convectional cobalt silicide formation process, which induces a rough interface due to Co and Si interdiffusion [14,15]. Besides, in this case, if the annealing temperature is lower than 750 °C, high resistive phase CoSi will be formed [14,15]. Comparing Fig. 6-3 and 6-5, we find that if the annealing process is well-controlled, a smooth interface and dense bulk cobalt silicide can be formed, which is the characteristic of oxide-mediated silicidation process [8-11] as in Fig. 6-3. However, over-annealing can destroy the cobalt silicide quality as in Fig. 6-5.
Vantomme [16] and Pretorius [17] reported that if the Co effective concentration at the Cobalt silicide growth interface were low enough, this would lead to the biggest negative change in the free energy for the CoSi2 formation. In addition, the direct CoSi2 formation can effectively reduce the formation temperature because the reaction path is Co + 2 Si Æ CoSi2 rather than CoSi + Si Æ CoSi2, in other words, it
needs not to break the CoSi bonding [18]. Therefore, a faster reaction rate of Ti with SiOx will enhance the Co diffusion rate and Co effective concentration at the cobalt silicide growth interface, which then induces high resistive CoSi formation. The control of the reaction rate between Ti and SiOx is crucial in oxide mediated silicidation with a Ti-capping process. Following the argument, whereas a higher reaction rate can increase cobalt silicide thickness but more easily form high resistive CoSi, a lower reaction rate can ensure CoSi2 as the first and only phase but reduce thickness. Besides, the SiOx layer must be remained after annealing for a dense bulk cobalt silicide film with a smooth interface between cobalt silicide and Si. Of course, the original thicknesses of Co, Ti and SiOx layers must be optimized before annealing to yield a high quality CoSi2 thin film.
6.4 Conclusions
The fact that a Ti-cappng layer can increase cobalt silicide thickness in oxide-mediated silicidation process is because Ti can absorb oxygen from the SiOx
layer and then dissipate or completely decompose the SiOx layer. Proper control of the reaction rate between Ti and SiOx during annealing to maintain the SiOx layer of some thickness for a diffusion barrier will form a high quality CoSi2 thin film of enough thickness. In addition, the remained SiOx layer also serves to prevent the direct
reaction of Co with Si, which then ensures a dense bulk cobalt silicide film with a smooth interface between cobalt silicide and Si.
Chapter 7
Formation of Pyramid-like Nanostructures during Cobalt Film Growth by Magnetron Sputtering
7.1 Introduction
Cobalt possesses superior electrical and magnetic properties and hence has attracted much attention on its fundamental growth and potential application [1,2].
Especially, cobalt nanostructures might be applied in magnetic recording media or carbon nanotube (CNT) growth as the most active catalytic site [3]. Metal with a sharp tip can also be applied in the field-emission flat panel displays [4-6] where the sharpened tip results in lower work function to give rise to high current density.
However, the pyramid-shaped tips often require an extra etching process [4-7] on cobalt films. In this paper, we found that pyramid-like nanostructures with sharp tips can be naturally formed without any post-etching process by selective growth with dc magnetron sputtering. The evolution of the pyramid-like nanostructure was systematically examined with various processing parameters and the growth mechanism is investigated. Through the study, we provide a simpler growth method to fabricate a sharp-tipped Co metal film.
7.2 Experimental Details
A dc magnetron sputter was employed to deposit Co thin films on n-type Si(100) and (111) substrates. The substrates were chemically cleaned before loading into the chamber. The Co target (99.95% in purity) was pre-sputtered for 10 minutes as soon as the base pressure of 3 x 10-6 torr was reached using argon (99.995% purity) as the sputtering gas. During deposition, the total gas flow was maintained at 40 sccm while other deposition parameters were varied to examine the microstructure evolution of the growing Co thin films. The experiment was performed at room temperature. On the characterization side, the surface morphology, texture and phase of the evolving thin films were examined by scanning electron microscopy (SEM) (PHILLIPS XL-FEG) operated at 15KV, X-ray diffraction (XRD) (RIGAKU D-MAX-IV) with Cu Kα radiation and transmission electron microscopy (TEM) (HITACHI HF-2000) operated at 200 keV, respectively. The TEM cross-sectional thin foils were prepared by focus ion beam (FIB) (FEI 205) operated at 30 kV with a Gallium source to about 0.1 µm in thickness.
7.3 Results and Discussion
Fig. 7-1 is the SEM images of the surface morphology of the as-deposited cobalt thin films with various substrate biases under the deposition distance of 6 cm and the applied power of 50W. The pyramid-like nanostructures are found to form on the
growing surfaces. It is worth noting that the pyramid-like nanostructures only occur under an appropriate negative bias range from -30 to -60 V, while they would not form under either positive bias or negative bias outside the aforementioned range. The nanostructures have well defined shapes with a sharp tip determined by minimum surface energy. The typical base width of the pyramids increases from about 100 nm at -30V to 250 nm at -50V followed by a decrease to 100 nm at -60V. No faceted nanostructures form with a substrate bias beyond this range. This certainly implies that the formation of nanostructures is related to the energy of the deposited adatoms, which is varied by the negative bias. Since the nanostructure formation is related to the energy, we examine the morphology evolution as a function of applied power from 40 to 70 W as shown in Fig. 7-2 for plan-view images and Fig. 7-3 for cross sectional images. The biggest size of the pyramid-like nanostructure is about 250 nm and occurs at the power of 50 W, while the size distribution changes from 60-220nm for 40 W to 20-150nm for 70 W. With increasing applied power, the average size decreases and density increases, indicating that ion bombardment effect is more pronounced to eliminate the nanostructure growth at high energy regime. Both experiments from varying substrate bias and applied power reveal that both adatom kinetic energy and etch by ion bombardment determine the nanostructure formation.
Fig. 7-4 is a bright-field TEM cross-sectional image and a diffraction pattern of a
pyramid-like nanostructure. The image shows that the nanostructure is a columnar
grain with hcp phase, which grew out of the Si substrate on a basal plane of 10 1 0.
The hcp structure is the same as the results from Co thin film deposition without substrate bias where the major Co phase changes from ε-Co (hcp) to α-Co (fcc) when deposition distance changes from 6cm to 10cm as evidenced from the diffraction patterns shown in Fig. 7-5. According to Thornton’s structure-zone model [8] for sputtered metal deposites, when the Ts/Tm value is between 0.1 and 0.4, where Ts (K) is substrate temperature and Tm (K) is melting point, the metal thin film will develop a typical characteristics of columnar grains for zone T microstructure. Under the lower deposition temperature condition, adatom has insufficient mobility to proceed a long lateral bulk diffusion so that grain growth develops into a columnar structure. Because the deposition was performed at room temperature and cobalt melting temperature is 1495ºC, the Ts/Tm value is then 0.17, which should correspond to the zone T microstructure. Two types of the pyramid-like nanostructures are found on Co thin films with various faceted planes shown and assigned in Fig. 7-6. Base on the angles
between individual faceted plane and the basal plane of 10 1 0, and the length ratios of one side to height of the basal plane triangle as sketched in Fig. 7-6(c), we find out
that type I is composed of 1 013, 02 21 and 2 201, and type II is composed of 1 013,01 1 1 and 1 1 01 [9]. In a hcp structure, (0002) plane normally acts as the habit
plane due to the lowest surface energy [10] and found to be established easily when adatoms can diffuse in large distance laterally with sufficient mobility [11]. When the (0002) plane develops, faceted planes disappear. This can explain why nanostructures can not be produced with higher applied power or substrate bias. Another evidence from texture evolution in the films by XRD is shown in Fig. 7-7, where the intensity ratio represents the percentage of the peak intensity from a specific plane to the intensity summation from all four major planes. Fig. 7-7 shows that preferred orientation is varied largely by substrate bias. The preferred orientation changes from
(10 1 0) and (11 20) to (0002) as increasing substrate bias, where the strongest (10 1 0) texture occurs at -50V corresponding to the highest possibility to produce Co nanostructure, consistent with the previous results. From Fig. 7-7, apparently, when the bias is larger than – 75 V, adatoms have enough mobility to form stable (0002) plane. Conversely, when the bias is less than – 75 V, other higher surface energy planes develop. The fact that less stable planes dominate the texture in the as-deposited film is often seen in the zone T structure and has been well explained by anisotropies in surface diffusivities and atomic shadowing [12]. Of course, the texture of a metal thin film is also influenced by deposition method, nature of the substrate, energetic ion bombardment, and geometrical confinement by surface features [13].
So far, we have discussed the conditions for Co pyramid-like nanostructure
formation but why a columnar grain can form a pyramid-like island on top? The experimental results in that the Co nanostructure would not form under zero or smaller negative bias even with the same applied power suggest that bias-induced ion sputtering also plays an important role in the formation. Ion sputtering have been shown to induce various surface topography, such as cone, pit and faceting etc. [14-16]
Regarding possible mechanisms responsible for pyramid formation upon ion sputtering, enhanced local erosion rate due to defect generation has been proposed [17-19] to explain the formation of symmetrical pyramids. However, both symmetrical and asymmetrical pyramids produced simultaneously with equal concentrations from our results imply that bias-induced surface diffusion to reach minimum surface energy configuration could be the major cause.
According to Thornton [20], when the deposition temperature is lower than Tm/3, intrinsic stress develops because of the low adatom mobility. Wolf [21] has also reported that sputtered metal films deposited on Si substrates at low substrate temperatures tend to exhibit tensile stress and the stress increases with thickness.
Because the pyramid-like nanostructure comes from a columnar grain structure, its grain growth process is a type of secondary grain growth [22]. Different to bulk grain growth, stress normally plays a significant role in the secondary grain growth process.
In order to understand the influence of stress in the grain growth, the surface
morphology evolution of Co thin films were examined as a function of film thickness under deposition distance of 6 cm, applied power of 50 W and substrate biases of –50 V as shown in Fig. 7-8. Fig. 7-8 reveals that the density of smaller-sized nanostructures decreases and that of larger-sized nanostructures increases with thickness. In the low-temperature thin film growth regime, bulk diffusion of adatoms does not occur, therefore, coarsening is slower and proceeds through grain boundary migration [23]. As a result, as increasing film thickness, stress increases and grain growth is enhanced by grain boundary migration, which then results in increasing nanostructure size.
We also study the surface morphology of fcc cobalt thin films deposited using 10 cm deposition distance (please refer to Fig. 7-5) with various powers from 60 to 90 W as shown in Fig. 7-9. Surprisingly, no nanostructures are formed on the surface for all conditions and the surface becomes smooth with increasing applied power. This demonstrates that the pyramid-like nanostructure can only be formed when the cobalt thin film is hcp phase.
In order to visualize substrate effect, cobalt was also deposited on Si(111) substrates using applied power of 75 and 125 W as shown in Fig. 7-10. The results show that the shape of nanostructures becomes more complicated, where pyramid-like nanostructures with pentagon or various polygon bases are observed. Nanostructures
disappear when the ion bombardment become more serious at higher applied power analogous to the case in the Si (001) substrate. These results suggest that substrate
disappear when the ion bombardment become more serious at higher applied power analogous to the case in the Si (001) substrate. These results suggest that substrate