Interfacial properties of [Pt/Co/Pt] trilayers probed through magnetometry

The magnetic and interface properties of [Pt/Co/Pt] were investigated. First, the magnetic properties were determined from the magnetic dead layer plots, in which the Co layer was considered as two distinct parts representing different magnetic properties. The two parts with low and high tCo ranges are close to and away from the top interface (Co/Pt), respectively. The part close to the top interface shows a smaller magnetization (M) value and nonlinear behavior. However, the other part shows a higher M value closer to the bulk value and a linear behavior. The nonlinear behavior of the M values of the low tCo range was converted to an impurity level using simple assumptions. The results showed the effect of the top Pt layer on the magnetic properties of the Co layer. The results clearly demonstrate that magnetometry could be utilized as a means to understand the interface quality of magnetic multilayer systems.

MDL plots. Consequently, a systematic magnetometry study was performed, which mainly involved plotting the magnetic moment as a function of t Co . These results are shown in Fig. 2a,b over a wide t Co range for the samples with t Pt = 0.25 and 3.0 nm, respectively. The upper set of results in each figure represents for the as-deposited samples, whereas the lower set represents the annealed samples. In these figures, not the magnetic moment itself but its normalized value according to the sample area (emu/cm 2 ) is plotted as a function of t Co so that the slope corresponds to the magnetization value. An obvious analytical equation describing the results in the high t Co range is y = ax + b . Coefficient a is identical to the saturation magnetization value (M s in emu/cm 3 ), and the x value at which y = 0 (viz., − b/a) indicates the magnetic dead layer (MDL) thickness. Notably, the MDL thickness obtained in this way is of little physical significance because the magnetic moment is not zero at the MDL thickness. This linear behavior indicates that the M s value can be determined in this t Co range. For the samples with t Pt = 0.25 nm, M s = 1379 emu/cm 3 for both samples in the as-deposited state and after annealing. However, for the samples with t Pt = 3.0 nm, this M s value is reduced substantially to 1333 emu/cm 3 in the as-deposited state, even though it increased slightly to 1355 emu/cm 3 after annealing. All these values are lower than those reported for  A deviation from the linear behavior is visible in the low t Co range. As the t Co value decreases, all the plots show an upward deviation, which contrasts with the conventional MDL plot showing a linear behavior. Considering that the slope of the plot is identical to the M s value, the upward deviation indicates that the M s value in this low t Co range decreases with decreasing t Co . This could be due to the interpenetration of nonmagnetic Pt atoms into the Co layer. A similar upward deviation was not observed in the MDL plots for [Pt/Co/Cu] trilayers 15 , indicating that the saturation magnetization of Co is constant in the Co thickness range considered in the MDL plots. As the relative portion of the interpenetrated region over the entire Co layer will increase with decreasing t Co , the observed results show a lower M s value at a lower t Co . For quantitative analysis, significant results could be obtained using an analytical equation. Although the choice for the analytical equation in the low t Co range is not so obvious, the following equation accurately describes the results: y = a(x − b) c , where parameters a, b, and c are summarized in Table 1. In this equation, exponent c denotes the deviation from the linear behavior. When c = 1 , this equation converges into its linear form for the high-t Co region. The extracted c values were very close to 1, indicating that the deviation from the linear behavior is not large. For the as-deposited samples, Magnetic properties in Co layer derived from MDL plots. Based on the analytical equations, magnetization (which is identical to the derivative of the equation, according to t Co (x)) can be obtained. Figure 3a shows the results for M (magnetization) as a function of the position in the Co layer. The schematics of the stack structure (left) and a typical variation of M with the position in the Co layer (right) are shown in Fig. 3b. For the bottom Pt/Co interface, where the intermixing level during sputtering is negligible, leading to a well-defined interface, the M value is close to the bulk value 11 . Notably, for all the cases, M ≠ 0 at t Co = 0, indicating the existence of an interface magnetization. A probable reason for the interface magnetization is that nonmagnetic Pt atoms can have a magnetic moment when they are in contact with magnetic Co atoms; this is known as the    18 reported that the induced magnetic moment does not differ significantly between the Pt/Co and Co/Pt interfaces; however, Kim et al. 19 reported that the proximity effect from the top interface is stronger than that from the bottom one. In this paper, no evidence is observed on the relative strength of the proximity effect between the bottom and top interfaces. Therefore, only the amount of proximity effect is considered. The interface magnetization values were not small, i.e., the values were 301.7 emu/cm 3 (for t Pt = 0.25 nm) and 328.8 emu/cm 3 (for t Pt = 3.0 nm) for the as-deposited samples and 174.5 emu/cm 3 (for t Pt = 0.25 nm) and 143.5 emu/cm 3 (for t Pt = 3.0 nm) for the annealed samples. These interface magnetization values due to the proximity effect can be converted into the magnetic moment possessed by one Pt atom magnetized using the following simple relation: The magnetic moments obtained are in the range of 0.25 to 0.54 μ B , which agree with the reported values 16,17 .

Estimation of impurity profiles in Co layer. Based on the magnetometry results and their analysis, a
schematic showing the concentration of interpenetrated impurity atoms in the Co layer can be drawn as a function of its location if the following two simplifying assumptions are made. First, intermixing during sputtering occurs only at the Co/Pt interface. Second, for the samples with t Pt = 0.25 nm, in addition to the Pt atoms, the Ru atoms (the capping layer), deposited on the top Pt layer, can penetrate the Co layer; this is likely as the Pt thickness is low. In this case, the penetration of Ru atoms is assumed to occur during the deposition of a 2.75 nm Ru capping layer so that the total thickness (3 nm) affecting the interpenetration can be identical to that of the sample with t Pt = 3 nm. While converting the results of M into the impurity concentration, it is necessary to have information on the variation of M s with respect to the concentration of the impurities (Pt and Ru); this has been detailed in [20][21][22] for Pt and in 23 for Ru. A simple linear assumption in the required composition range is considered reasonable for the following two reasons. First, Co can have an fcc structure for a very thin Co layer. Therefore, Co atoms are likely to be miscible with Pt with an fcc structure [24][25][26] . Second, the Co-Ru binary phase diagram indicates that the atoms are miscible in the range up to 6 at.% of Ru. An equation used for the conversion is as follows.
Here, M s,Co and M s,sample are, respectively, M s values of a pure Co and the Co layer interpenetrated with impurity atoms. A imputiry indicates the decrease of M s with an addition of 1 atomic percent impurity, whereas C impurity denotes the impurity composition. Figure 4 shows the calculated results for the concentration of interpenetrated impurity atoms in the Co layer. The results are related to the position in the Co/Pt interface. As observed in Fig. 4, the concentration of interpenetrated impurities in the Co layer is not small. At the interface, the impurity level for the samples with t Pt = 3.0 nm is as high as 37.2 at.% (as-deposited) or 43.7 at.% (annealed). The impurity level is lower for the sample with t Pt = 0.25 nm, which is 14.5 at.% (as-deposited) or 16.  www.nature.com/scientificreports/ performed a detailed analysis in this study, as our main concern is to know the level of impurity and its change upon annealing. As expected, the concentration of impurities decreases monotonically as it is located away from the interface. For example, at a position of 0.5 nm, which is relevant to the t Co value in [Pt/Co] multilayers with the inverted structure 4,5 , the impurity concentration is 23 at.% for the samples with t Pt = 3.0 nm. In contrast, at the same position, the impurity concentration is estimated to be 8 at.% for the samples with t Pt = 0.25 nm, which is an optimum Pt thickness in the inverted [Pt/Co] multilayers. This explains the deteriorating effect of the interpenetrated Pt atoms on the PMA strength of the [Pt/Co] multilayers.

Discussion
The effects of the top Pt layer thickness and annealing on the interface quality of [Pt/Co/Pt] trilayers were systematically investigated. Even with the cross-sectional HRTEM, it is difficult to identify the exact location of a very thin layer such as t Pt = 0.25 nm. However, the HRTEM images, as shown in Fig. 1a-d, clearly show that the layer forms a continuous structure and its interfaces are atomically flat. These features are duly reflected by the magnetic properties and in this sense, the magnetometry can be a good tool to examine the interface properties of ultrathin magnetic films. From the MDL plots, the Co layer can be broken down into two parts that show different magnetic properties. In the high t Co range, a linear behavior was observed. However, a nonlinear behavior was observed in the low t Co range. Further, the proximity effect was detected in the low t Co range. The nonlinearity in the M values can be converted to the Co-layer impurity concentration by using an analytical equation. The interpenetration and inter-diffusion depth were found to be sensitive to t Pt and the annealing process. Specifically, for the samples with t Pt = 0.25 nm, the impurity concentrations near the top Co/Pt interface are significantly smaller than that for the samples with t Pt = 3.0 nm. This explains the relationship between the interface quality and PMA strength of the [Pt/Co] multilayers system. Although the impurity levels can vary depending on the assumptions made earlier, the relative impurity level is minimally affected by these assumptions. Therefore, the magnetometric investigation of the interfacial properties will aid in analyzing the interface quality of the magnetic multilayers system.

Methods
The stack structure examined in this study consisted of the following: Si substrate (wet-oxidized)/Ta (5 nm)/Pt (10 nm)/Ru (30 nm)/Pt (3 nm)/Co (t Co )/Pt (t Pt )/Ru (3 nm). The two variables were t Co (the thickness of the Co layer between the two Pt layers) and t Pt (the thickness of the Pt layer on top of the Co layer). Thickness t Co varied between 0.5 and 10 nm, whereas t Pt was fixed at 0.25 or 3 nm. The samples were fabricated using an ultrahigh vacuum magnetron sputtering system with a base pressure of 8 × 10 −8 Torr. All the layers were deposited at a constant Ar pressure of 2 × 10 −3 Torr. No specific substrate cooling or heating was applied during the sputtering process. The thicknesses of the constituent layers were measured using a surface profiler. The deposition rate of the layers was adjusted to ~ 0.03 nm/s by varying the sputtering power. This deposition rate was used to calculate the thicknesses of the layers. The samples were annealed at 400 °C for 1 h under a vacuum pressure of 1 × 10 −6 Torr. The magnetic moment was measured using a vibrating sample magnetometer (VSM), and the microstructure was examined using high-resolution transmission electron microscopy (HRTEM).