52
53
fixed working distance (80 mm) and rotation speed (15 rpm). In one-component system, many factors are involved in the change of the deposition rate. At a fixed Ar pressure, working distance and rotation speed, the form of power supply, and the setting power are the important factors in influencing the deposition rate. In addition, a higher deposition rate occurs using a DC power than a RF power. Furthermore, a different setting power would bring about the several variations between the plasma and the substrate, such as the quantity of the incident atoms, the mean free path of the incident atoms, the ratio of the adhesion of the incident atoms and desorption for the as-deposited atoms, during the deposition process.
First, the power of cathodes lead to the quantitative variation of the incident atoms and the change of average mean free path, directly proportional to the inverse probability of collision, as shown in Figure 5-2 [52]. The curve can be divided into three regions. Region I exhibits a low probability of collision, implying a low deposition rate due to a very low plasma density. In region II, the increasing plasma density would cause more and more incident atoms with the higher kinetic energy forms and the probability of collision increases with increasing volt, meaning that a higher deposition rate appears in this region. Then, the plasma density reaches a saturated situation in region III, but a higher power would give the incident atoms the more kinetic energy, implying the probability of the collision would decreases again. However, the adhesion and desorption of the incident atoms are strongly affected by the kinetic energy of the incident atoms. A very high power would lead the as-deposited atoms to be “re-sputtering” by the incident atoms. Hence, the deposition rate becomes lower again.
As mentioned above, the several important factors involved in the deposition rate in one-component have been discussed. In the binary co-sputtering system, the competition between two elements plays an important role during the co-sputtering process. Above all,
54
compared with Mg and Cu, Mg exhibits a large radii (1.62 A。) and low atomic number (24), but Cu exhibits a small radii (1.28 A。) and large atomic number (64). According to these known data, it can be infer that Cu incident atoms, like small but heavy bullets, with a high deposition rate bombard the substrate at a high Cu power, leading to a loose structure, even micro-pores. Relatively, Mg incident atoms, like balloons, with a slow deposition rate land on the substrate, resulting in a compact structure. The average free path of Cu decreases owning to the participation of Mg incident atoms with a slow deposition rate. However the increased probability of collision between Mg and Cu would lead the transformation of the kinetic energy and momentum from Cu to Mg, leading to the overall increase of surface diffusion ability. During the co-sputtering process, the collocation of DC Cu and RF Mg deposition provides a good mixing between Mg and Cu, small Cu atoms with a high kinetic energy and momentum.
According to Figure 5-1, the irregular variation of composition occurs, suggesting it is mainly caused by “re-sputtering” between the Mg and Cu elements in the co-sputtering system. “Re-sputtering” means partial as-deposited atoms are re-sputtered from the substrate by the incident atoms. In the case of 100 series, the Cu contents are approximately close to the expectable values, calculated from the data of the isolated Mg and Cu depositions. Then, compared with the 100 series, the Cu content of the 50 series specimens are all higher than that of the 100 series since the large difference in the kinetic energy between Mg and Cu. The incident atoms of Cu with a relatively high kinetic energy would impact the surface of the film and make Mg atoms escape from the surface at low Mg powers, such as 50-150 and 50-100, showing the absence of Mg. Hence, the difference between the kinetic energy of Mg and Cu is the key which leads to the deviation of the composition.
The slight negative heat of mixing in the Mg-Cu system would lead to the Mg-Cu
55
clusters while the incident atoms deposit on the substrate, as well as some Mg-Mg and Cu-Cu clusters, too. Incident atoms with high kinetic energy can break up the Mg-Mg and Cu-Cu clusters by implanting to form the more stable Mg-Cu clusters. In the case of low Mg and Cu powers, insufficient kinetic energy can not destroy the clusters of pure Mg and Cu, resulting in separated Mg and Cu nano-grains.
To summarize, the competition between the momentum and kinetic energy of different kinds of incident elements play a critical role in the co-sputtering procedure. The compositions of the 100 series obtained were close to expectation while the Cu power is higher than 25 W. However, a possible reason, re-sputtering, would make the abnormal variation of the compositions in the 50 series. Hence, the composition of the Mg-Cu thin films, from Mg17.7Cu82.3 to Mg61.9Cu38.1, can be easily controlled at RF target of Mg at 100 W and the DC target of Cu from 25 to 150 W.
5-2 Oxidation of Mg-Cu co-sputtered film
During the deposition process, the incident atoms with energetic energy and momentum continuously attach to the formed film on the substrate, and the kinetic energy and momentum transformation from incident atoms to as-deposited atoms would make the surface diffusion to find the most stable atomic sites as the residual kinetic energy is enough.
In other words, sputtering is a special process with surface diffusion, meaning a real-time heat-treatment occurs at a relatively low temperature near the free surface, compared with melt-spinning. Then, in the Mg-Cu co-sputtering process, the content of oxygen contained in the sputtered films suddenly rises about 15 at% while the Cu power is lower than 25 W, as shown in Table 4-1, suggesting that predominant Mg atoms react with oxygen in the air.
56
In the case of the 100 series, the proper excess energy causes some Mg-Cu crystalline clusters to overcome the energy barrier to transfer to the amorphous clusters or to form the Mg-Cu compounds. However, in the case of low Cu powers, such as 100-25 and 100-15, as shown in Figure 4-1, the as-deposited specimens contain the separated nano-grains of Mg and Cu with their specific structure respectively. A possible kinetic reason is suggested as the formations of intrinsic Mg-Mg, Cu-Cu, and Mg-Cu clusters, while Mg and Cu atoms adhere to the surface of the film during sputtering. The clusters of Mg-Mg, Cu-Cu, and Mg-Cu should disperse in random. At this moment, with insufficient kinetic energy and momentum, most of as-deposited atoms freeze on the free surface. In this case, Mg would rapidly react with oxygen while the film is exposed to air.
From the thermodynamic aspect, the heats of mixing among Mg, Cu, and O are the major factors influencing the formation of oxide or intermetallic compounds. Compared with the heats of mixing of MgO, Cu2O and the Mg-Cu solid solution, the heats of formation of MgO, Cu2O, and the heat of mixing of Mg-Cu solid solution are equal to -601.6, -167.5 and -3 kJ/mol at 298 K, respectively, meaning that the formation of oxide in the atmosphere with sufficient oxygen is very easy, especially for MgO. In other words, it is reasonable to suggest that the bonding between Mg and O is very easy to form comparatively.
According to the measured oxygen content, Cu atoms around Mg atoms can reduce the activity of Mg, avoiding oxidation of Mg in air. The reason for incomplete mixing is the incident atoms with the insufficient kinetic energy can not cause the surface diffusion to mix well, even induce the nanocrystalline Mg and Cu which can rapidly react to oxygen.
5-3 Diffusion-induced phase transformation in multilayered thin films
57
In the Mg-rich multilayered case, the diffusion starts at a relatively low temperature about 363 K. Before this experiment, Arcot et al. [62] and Rodriguez-Viejo et al. [63] have done the related research on the Mg-Cu multilayered films. In their researches, the composition ratio were set to 1(Mg):2(Cu) and 2(Mg):1(Cu). Regardless of the temperature (370 to 773 K) and composition in their studies, the first intermediate phase is always Mg2Cu.
Nevertheless, Mg2Cu does not form rapidly but step by step. The lateral growth occurs as the initial stage of the diffusion of Mg. Then, the perpendicular growth occurs with a continuous Mg2Cu layer. Hence, an experiment of Mg-Cu multilayered thin films is designed to compare the thickness effect in the individual layers.
First, the heat treatment of 20T32, consisting of 10 layers of 150-nm-thick Mg layer and 10 layers of 50-nm-thick Cu layer, was subjected to annealing at 423 K under 5x10-2 torr.
According to the XRD result, as shown in Figure 4-17, the Mg2Cu phase formed after 30 minutes. As the time of the heat treatment increases, the content of Mg2Cu increases greatly.
Then, another multilayered film designed as the Cu content shift to about 70%, or the 20T14, was annealed at 363 K. However, the interaction between Mg and Cu layers still exhibits the similar result of 20T32 mentioned above, as shown in Figure 4-18. Mg2Cu gradually forms as the annealing time increase.
Generally speaking, the metastable phases fabricated by the rapid quenching process is the amorphous phases. However, the different occurrences of the metastable phases in the multilayered system after post-annealing have been discovered, such as the formation of the FeB amorphous phase in the Fe/B multilayered structure after post-annealing [67] and the absence of the NiMn2 and Ni2Mn compounds during the Ni-Mn diffusion couple [57].
The possible reasons that diffusion-induced metastable phase in Mg-Cu multilayer
58
system does not form during low-temperature heat treatment are summarized below.
According to previous researches [62,68,69], the dominant factors of the formation of Mg2Cu are the element with a lower melting point and the structure of the compound. In other words, Mg with a lower melting point should dominate the diffusion. At an enough temperature, Mg atoms near the interface start to diffuse. However, the solubility of Cu atoms in the Mg matrix or Mg atoms in the Cu matrix is almost zero. Hence, the thin Mg2Cu layers form near the interfaces at the initial stage of the annealing. Once the intermediate layers of Mg2Cu form, more Mg2Cu nucleate and crystallize along these layers.
Also, Arcot et al. [62] reported that the transformation from Mg2Cu to MgCu2 at temperatures higher than 548 K in the Cu-rich alloy. In other words, MgCu2 is the high-temperature phase. Moreover, they also pointed out the first phase formed in the Cu-rich alloy is still Mg2Cu. With increasing temperature close to the liquidus line of Mg2Cu equal to 758 K, another transformation occurs from Mg2Cu to MgCu2 by adding three Cu atoms. The overall reaction sequence in films with an atomic concentration ratio of 2(Cu):1(Mg) is shown below,
2 K
548
2Cu 3Cu 2MgCu
Mg Cu 4 Mg
2 + → + ⎯⎯→ . (5-1)
In the thick film, the thicknesses of the individual layers do not affect the formation of the intermediate phase due to the poor solubility of Mg in Cu and Cu in Mg. However, while the thicknesses of the individual layers are lower than a critical value, the reaction in the multilayer would accelerate due to the increase of the interface energy of Mg-Cu multilayer films. Figures 4-19 and 4-20 exhibit the results of the heat treatment of 40N32 and 40N14. It is obvious that Mg and Cu peaks of 40N32 vanish after 30 minutes at 423 K, as well as
59
Mg2Cu formed much rapidly, suggesting that size effect accelerates the reaction of Mg and Cu. Then, 40N14, the Cu-rich multilayer, annealed at 363 K, exhibits the similar occurrence.
If the thickness of the individual layer is lower than 5 nm, it can be postulated that the microstructure would be similar to 50-50. Due to very high interface energy, the diffusion-induced reaction between Mg and Cu would occur at the temperature lower than 363 K.
Concluding the previous research and these experiments, the first intermediate phase of the Mg-Cu multilayered film is always Mg2Cu due to the localized diffusion around the interface. No other Mg-Cu intermediate phase, such as the Mg-Cu amorphous phase, was found during low-temperature annealing.
5-4 Comparison between Mg-rich and Cu-rich amorphous alloys fabricated by sputtering and liquid-quenching process
According to previous researches of Detendler et al. [70], the glass-forming composition of Mg-Cu amorphous alloy were calculated using the thermodynamic theory. In their study, Mg2Cu with a complex orthorhombic structure is assumed not to compete with the Mg-Cu amorphous phase to simplify the difficulty in calculation during liquid-quenching. However, MgCu2 with a simple face-centered cubic structure would compete with the Mg-Cu amorphous phase to induce the crystallization during liquid-quenching. Hence, the Mg-Cu amorphous alloy is believed to be different to form for a Cu-rich composition. Finally, the calculated glass-forming composition range is located Mg1-xCux ( 0.14 < x < 0.23 ) [70], for example Mg20Cu80. Moreover, in addition to MgCu2, the major reason why Cu-rich amorphous alloy is hard to form via liquid-quenching is that the solidus line in a Cu-rich composition is much higher than in a Mg-rich composition, leading to a higher processing
60
temperature which implies the Cu-rich amorphous alloys need a much higher cooling rate. A large difference between the melting points of Mg and Cu is the key influencing factor for the formation of amorphous alloy via liquid-quenching. Experimentally, Mg-Cu metallic glasses can be fabricated using melting-spinning in Mg1-yCuy ( 0.11 < y < 0.22 ) [71], for example, Mg85Cu15.
However, the above argument becomes unsuitable for the sputtering system. There are some points different from the liquid-quenching process:
(1) Sputtering is a low-temperature process near room temperature compared with the quenching process, meaning that the process temperature in sputtering is much lower than melting-spinning.
(2) The heat gradient for the active region of the gas-solid phase transformation in sputtering is almost zero, compared with for the liquid-solid phase transformation in melt-spinning.
(3) Incident atoms are mixed with the deposited atoms near the surface continuously during sputtering, unlike atoms in the melting state freeze within the very short time during melt-spinning.
(4) The structure of the substrate influences the structure of the as-deposited films.
Due to the four different points, the properties of the specimens fabricated using the sputtering would differ from the specimens fabricated using the liquid-quenching process. It is discussed below for the discrepancy between the Mg-rich and Cu-rich amorphous alloy via the sputtering and liquid-quenching processes.
In the study of Ong et al. [72] in 2001, the XRD pattern of the as-spun Mg79Cu21
fully-amorphous ribbon by melt-spinning exhibits a hump from 33o to 43o as shown in Figure
61
5-3. Then, the XRD pattern of the as-deposited Mg61.9Cu38.1 thin film by sputtering exhibits a hump from 33o to 44o containing Mg2Cu particles, as shown in Figure 4-19, which structural transformation during vacuum annealing is very similar to Mg79Cu21. In short, Mg-rich as-spun and as-deposited specimens exhibit a similar transformation of Mg2Cu during annealing.
As mentioned above, the Cu-rich amorphous alloy is hard to form due to a necessary high cooling rate. Hence, the Cu-rich amorphous alloy via the liquid-quenching method has not been reported. Kaya and Smith [73] fabricated the Mg9.7Cu91.3 alloy via rapid quenching with an insufficient cooling rate in 1992, resulting in a Cu-MgCu2 lamellar structure, as shown in Figure 5-4. However, sputtering with a high cooling rate indeed makes a possibility to form the Cu-rich amorphous alloy in this study, as shown in Figure 4-1. The high-resolution image of the Mg23.5Cu76.5 film deposited on the Si (100) wafer, as shown in Figure 4-12, reveals the MgCu2 nanoparticles embedded in the Mg-Cu amorphous matrix.
Additionally, Table 5-1 exhibits the thermal properties of Mg-Cu binary and Mg-Cu-X (X = Y and Gd) ternary metallic glasses. Compared with the Mg-Cu-X (X = Y and Gd) amorphous alloys, a low ΔTx value of the Mg23.5Cu76.5 alloy represents the ability to resist crystallization during heating due to the lack of the third element to suppress the formation of MgCu2 with a simple FCC structure.
Concluding speaking, sputtering is a potential process to fabricate thin film metallic glasses. However, two points of conclusions can be reached about the reasons for the difficulty in forming fully amorphous thin films using co-sputtering, as listed below.
(1) In addition to the mentioned factors during sputtering, such as the working pressure, working temperature, power of cathode, etc, the structure of the substrate plays a very important role in inducing the variation of the as-deposited film structure. The
62
Cu-rich co-sputtered films deposited on Si (100) wafers exhibit a partially amorphous structure. To improve the co-sputtering process, amorphous substrates, such as glass or polymer would promote the formation of fully amorphous Mg-Cu films.
(2) During co-sputtering, another important key is the power adjustment between Mg and Cu cathodes, avoiding the occurrence of the re-sputtering effect, the formation of MgCu2 with the simple FCC structure and the freezing of incident atoms with insufficient residual kinetic energy.
5-5 Nano-mechanical properties of 100-150 (Mg
17.7Cu
82.3), 100-100 (Mg
23.5Cu
76.5), and 100-50 (Mg
40.4Cu
59.6)
For metallic glasses, the specific characteristic pop-in effect is considered to reflect the plastic deformation contributed by shear bands. As shown in Figure 5-5 [74], the load-displacement curve refers that one pop-in of three parts, as shown below,
Δh=Δhfast +Δhslow +Δhe, (5-5)
where Δh, Δhfast, Δhslow, and Δhe is the overall displacement in one pop-in, the shear-band formation, the progressive plastic deformation, and the elastic deformation, respectively. In order to determine the correct pop-in size, the load-displacement curves , as shown in Figures 4-27, 4-28, and 4-29, must be modified to remove the influence of the elastic deformation (Δhe) by Oliver-Pharr relation [64], as expressed by,
P=B(h−hf)m, (5-6)
63
where P is the load applied to the test surface, h is the resulting penetration, B and m are empirically determined fitting parameters, and hf is the final displacement after complete unloading. Then, the unloading stiffness, S, is established by analytical differentiation, as shown below.
m f h
h S mB h h
dh
dP −max= = ( − ) . (5-7)
Finally, the contact depth can be estimated using the following equation, as shown below,
S h P
hc = l −η , (5-8)
where hc and hl are contact depth and the displacement in the loading part, respectively, and η is a constant depending on the indenter geometry. For Berkovich and Vicker indenters, η is 0.75.
As people know, the mechanical properties are influenced by the variation of the composition in the alloy systems, which would lead to the difference of the microstructure, especially in BMGs and the composites consisting of glassy and crystalline phases. For the co-sputtered specimens of 100-150, 100-100, and 100-50 indented at the strain rate of 5×10-3 s-1, the modulus-displacement and hardness-displacement curves are shown in Figure 4-25.
General speaking, the deviation of the composition in a binary system, such as the Cu-Zr system, would affect the microstructure due to the variation of the glass-forming ability[75].
Then, the difference of the microstructure also affects the mechanical and thermal properties.
However, in a simple binary system, such as Mg-Cu, is not suitable completely. In a simple
64
system, the deviation of the alloy composition maybe leads to the different quantity of Mg-Cu amorphous phase and MgCu2. Hence, the 100-150 and 100-100 samples exhibit similar thermal properties but dissimilar mechanical properties in the nanoindentation test.
According to Figure 4-25, the values of Young’s modulus and hardness in the initial stage show a few irregular points due to the geometry of the indenter. 100-100 exhibits a higher Young’s modulus and hardness in different indented depths, suggesting that the high Young’s modulus is majorly contributed by MgCu2. Similarly, the off-eutectic composition of the 100-50 specimen, locating in the Mg2Cu-MgCu2 eutectic region, lowers the value of the Young’s modulus. Figure 4-26 presents the comparison of Young’s modulus and hardness among the 100-150, 100-100, and 100-50 specimens.
It is known that an intermetallic compound typically exhibits the hard and brittle characteristics, possessing a very high Young’s modulus and hardness. In the 100-100 specimens, MgCu2 particles suggested be the major source provides the obstacle to constrain the movement of the shear band and finally lead to enhance the Young’s modulus.
As-mention above, in a simple binary Mg-Cu system, assuming that the amorphous structure of 100-150 and 100-100 is very similar, the composition difference would lead to the different quantity of the MgCu2 phase in the Mg-Cu amorphous matrix. In other words, the higher Young’s modulus of 100-100 than 100-150 could be due to the 100-100 specimen contains the more MgCu2 phase than the 100-150 specimen, implying 100-150 is more amorphous than 100-100. Then, the plastic deformation caused by shear bands would be contributed by the sub-structure, which will be discussed below.
Figure 5-6 exhibits the different “pop-in” effects in the 100-150, 100-100 specimens.
The size of the “pop-in” effect in 100-100 is smaller than those in 100-150 but more compact,