# Fileset

[RDE_manusscript_MVG_CPP-GMR_clean.pdf](https://mdr.nims.go.jp/filesets/583ef995-0d29-4529-ab88-ed61e6960878/download)

## Creator

[Hirofumi Suto](https://orcid.org/0000-0003-4387-5862), [Vineet Barwal](https://orcid.org/0000-0001-9445-5900), [Kodchakorn Simalaotao](https://orcid.org/0000-0002-6098-4422), [Zehao Li](https://orcid.org/0000-0002-3867-1462), [Keisuke Masuda](https://orcid.org/0000-0002-6884-6390), [Taisuke Sasaki](https://orcid.org/0000-0002-5952-7638), [Yoshio Miura](https://orcid.org/0000-0002-5605-5452), [Yuya Sakuraba](https://orcid.org/0000-0003-4618-9550)

## Rights

[Creative Commons BY Attribution 4.0 International](https://creativecommons.org/licenses/by/4.0/)

## Other metadata

[Negative spin polarization of Mn2VGa Heusler alloy thin films studied in current-perpendicular-to-plane giant magnetoresistance devices](https://mdr.nims.go.jp/datasets/efd3ecc1-f33f-489e-930f-16f3e7f3c39b)

## Fulltext

1  Negative spin polarization of Mn2VGa Heusler alloy thin films studied in current-1 perpendicular-to-plane giant magnetoresistance devices 2 Hirofumi Suto*, Vineet Barwal, Kodchakorn Simalaotao, Zehao Li, Keisuke Masuda, Taisuke 3 Sasaki, Yoshio Miura, and Yuya Sakuraba 4 1Research Center for Magnetic and Spintronic Materials, National Institute for Materials Science 5 (NIMS), Tsukuba, 305-0047, Japan 6 SUTO.Hirofumi@nims.go.jp 7  8 ABSTRACT 9   Magnetic materials with high negative spin polarization have been sought as a building block to 10 increase the design freedom and performance of spintronics devices. In this paper, we investigate 11 negative spin polarization of Mn2VGa Heusler alloy in current-perpendicular-to-plane giant 12 magnetoresistance (CPP-GMR) devices. We fabricated an epitaxial CPP-GMR stack consisting 13 of Mn2VGa/Ag/CoFe with L21 ordering in the Mn2VGa layer and observed negative 14 magnetoresistance (MR), which provided evidence of negative spin polarization. The MR ratio 15 depended on thermal treatments (deposition at an elevated temperature and post-annealing), 16 because these processes affected the ordering, roughness, and magnetic properties of Mn2VGa. 17 The maximum MR ratio reached −1.8% at room temperature and −3.0% at low temperature, 18 representing the highest among the negative MR values in pseudo-spin-valve CPP-GMR devices 19 despite the underestimation due to an incomplete antiparallel magnetization configuration. These 20 findings demonstrate the potential of Mn2VGa for a material with high negative spin polarization.  21  22   23 mailto:SUTO.Hirofumi@nims.go.jp2  I. INTRODUCTION 24   Material research has played a crucial role in improving the performance of spintronic devices. 25 One key research direction has been the search for materials with high spin polarization, because 26 these materials can significantly enhance magnetoresistance (MR) and spin-transfer torque (STT) 27 efficiency, which are the main operation principles of the spintronic devices  [1–7]. For example, 28 Co-based Heusler alloys, predicted to possess +100 % spin polarization, have been extensively 29 studied and have experimentally demonstrated very high MR ratios  [8–15]. On the other hand, 30 another research direction into negative spin polarization materials has also gained recent 31 attention. In negative spin polarization materials, the spin momentum direction of the spin-32 polarized conduction electron is opposite to the magnetization direction [16,17]. This relation is 33 in contrast to the standard positive spin polarization found in the aforementioned Co-based 34 Heusler alloys and many other spintronic materials. Intriguingly, negative spin polarization 35 materials can reverse the signs of MR and STT. In particular, negative STT can realize new device 36 structures by exerting STT of the desired spin direction and improve their performance beyond 37 the limitation of conventional positive spin polarization materials. This characteristic is 38 advantageous to devices demanding strong STT, such as spin-torque oscillators for energy-39 assisted writing in hard-disk-drive applications, in which oscillation of large magnetic volume is 40 crucial  [18–23]. 41  42   Tunnel magnetoresistance (TMR) studies on Fe4N and Fe3O4 have reported large negative MR 43 over −50% at room temperature, demonstrating the negative spin polarization of these 44 materials  [24,25]. Similarly, giant magnetoresistance (GMR) studies reported that materials such 45 as CoFeGd, FeCr, FeV, NiCr, and Fe4N have negative spin polarization [16,26–29]. However, 46 the amplitude of the reported negative MR ratio is smaller than −1% at room temperature in the 47 case of pseudo-spin-valve current-perpendicular-to-plane (CPP)-GMR devices, indicating the 48 small negative spin polarization of these materials. Considering that negative STT is beneficial to 49 3  devices demanding strong STT, implementation of negative spin polarization in CPP-GMR 50 devices is highly demanded for practical applications because their all-metallic structure can 51 withstand higher current density than TMR devices. 52  53   Mn2VGa (MVG) belongs to Mn-based Heusler alloys and is a promising candidate for a negative 54 spin polarization material  [30–36]. The calculated density of states has a pseudo gap in the 55 majority-spin band, from which high negative spin polarization is expected  [33–35]. In addition, 56 its Curie temperature of ~784 K is sufficiently high for room temperature operation. The 57 saturation magnetization (Ms) is small due to the ferrimagnetic configuration of the Mn and V 58 moments, and this property is suitable for hard disk drive applications because the small stray 59 fields from small Ms materials can minimize disturbance to the writing field  [22,37]. Klewe, et 60 al.  [32] fabricated TMR devices using a sputter-deposited epitaxial stack consisting of MgO(001) 61 substrate/MVG/MgO barrier/Co0.7Fe0.3 and confirmed negative MR, providing evidence of 62 negative spin polarization in MVG. However, no study has been reported on CPP-GMR devices.  63  64   In this study, we investigated the negative spin polarization of MVG in epitaxial CPP-GMR 65 devices. We fabricated GMR devices with MVG and CoFe magnetic layers with a Ag spacer and 66 demonstrated a negative MR ratio of –1.8% at room temperature. A tungsten layer provides a 67 suitable template that enables a high-quality epitaxial MVG layer to grow and stops the 68 interdiffusion between the MVG layer and the low resistance Ag bottom electrode. Deposition of 69 the MVG layer at an elevated temperature followed by post-annealing can enhance the ordering 70 and negative spin polarization of MVG while maintaining the flatness of the GMR film. These 71 results demonstrated the potential of MVG as a high-performance negative spin-polarization 72 material. 73  74 II. EXPERIMENTAL AND COMPUTATIONAL METHODS 75 4    We prepared the following three series of samples by magnetron sputtering on MgO(001) 76 substrates: (A) single-layer MVG films, (B) MVG films on bottom electrodes, and (C) CPP-GMR 77 stacks with an MVG layer. The numbers in the parentheses in the following are thickness in 78 nanometers. The series-A samples were composed of MVG (37)/Al (2), in which the MVG layers 79 were deposited directly on MgO substrates and passivated by an Al layer. The series-B samples 80 were composed of a bottom electrode/ MVG(37) / Al(2), where the  bottom electrodes were Cr(5) 81 / Ag(50), Cr(5) / Ag(50) / V(5), and Cr(5) / Ag(50) / Cr(5) / W(15). The series-C samples were 82 composed of Cr(5) / Ag(80) / Cr(5) / W(15) / MVG(10) / Ag(7) / Co50Fe50(7) / Ag(5) / Ru(8). The 83 MVG layers were deposited by co-sputtering from Mn55Ga45 and V targets at elevated substrate 84 temperatures (Ts) ranging from 300°C to 600°C. In some samples, the MVG layers were 85 additionally post-annealed at a temperature (Tp) of 500°C for 30 min after the deposition. The 86 layers above the MVG layer were deposited at room temperature after cooling down. The typical 87 composition of the MVG layer was Mn54.29V20.45Ga25.26, as measured by X-ray fluorescence (XRF) 88 analysis calibrated by the standard sample whose composition was analyzed by inductively 89 coupled plasma mass spectrometry. The MVG composition was adjusted to be Mn-rich and V-90 poor because this off-stoichiometric composition can achieve higher ordering  [36]. In addition, 91 anti-site of V atoms occupying the Mn site is detrimental to negative spin polarization according 92 to the calculation  [32], and a Mn-rich and V-poor composition can suppress such anti-sites  [36]. 93  94   The crystal structure and atomic ordering of the prepared films were investigated using X-ray 95 diffraction (XRD) with a Cu–Kα radiation source. The degree of B2 and L21 order (𝑆𝐵2 and 𝑆𝐿21) 96 were calculated as SB2 = √I002exp/I004expI002sim/I004sim  and SL21 = √I111exp/I444expI111sim/I444sim , where Ihklexpand 𝐼hklsim  represent the 97 experimental and simulated hkl peak intensities, respectively. In the XRD simulation, the off-98 stoichiometric composition was taken into account in the simulation model, where the excess Mn 99 atoms simply occupy the deficient V sites. The magnetic properties of MVG were evaluated by 100 5  measuring the curves of magnetization (M) versus magnetic field (H) using a vibrating sample 101 magnetometer at room temperature. The Ms was estimated from the M-H curves. The surface 102 morphology was measured using atomic force microscopy with a scan area of 1 μm × 1 μm, and 103 the average roughness (Ra) and the peak-to-valley height (P-V) were calculated. The cross-104 sectional microstructure of the samples was analyzed using high-angle annular dark-field 105 scanning transmission electron microscopy (HAADF-STEM), nano-beam electron diffraction 106 (NBED), and energy-dispersive X-ray spectroscopy (EDS).  107  108   Pseudo-spin-valve CPP-GMR devices were fabricated using the following procedure. The 109 series-C samples were patterned into circular and elliptical pillars with dimensions of 80 nm × 80 110 nm and 140 nm × 70 nm by electron-beam lithography and Ar ion milling. The pillars were then 111 covered with a thin Ta adhesion layer and a SiO2 passivation layer. After the lift-off of the 112 covering layers, a Au top electrode was fabricated. For each device size, 75 devices were 113 fabricated on one substrate. Resistance versus magnetic field (R-H) measurements were 114 conducted using the four-probe method. All the devices were measured by an auto-prober system 115 with an H range of 2.5 kOe at room temperature. A few devices were additionally measured by a 116 physical properties measurement system using a higher H range and lower temperatures. The in-117 plane magnetic field was applied along the longitudinal direction of the elliptical pillars. The MR 118 ratio was defined as 𝑀𝑅 = (𝑅 − 𝑅max) 𝑅max⁄ , where 𝑅max is the maximum R in the measured H 119 range.  120  121 The first-principles calculations for ballistic transmittance based on the Landauer formular  [38] 122 were performed using the QUANTUM ESPRESSO code  [39,40]. For the exchange-correlation 123 energy, we adopted the ultrasoft pseudopotential method and the Generalized Gradient 124 Approximation  [41]. A set of 10 × 10 × 1 grid of k-points is used for the Brillouin zone 125 integrations. The cutoff energy for the wave function and charge density is set to 40 (Ry) and 400 126 6  (Ry), respectively. Stacking models of MVG/Ag/MVG(001) were constructed using tetragonal 127 supercells, with the in-plane lattice parameter of the supercell fixed at a = 4.067 Å. The 128 MVG/Ag(001) interface featured two types of termination: Mn termination and VGa termination. 129 The stacking models consist of 7 atomic layers of Ag spacer, with 17 and 15 atomic layers of 130 MVG for Mn termination and VGa termination, respectively. 131  132 III. RESULTS AND DISCUSSION 133   We first studied the epitaxial growth of MVG and the effect of Ts on the ordering using the 134 series-A samples. Figure 1(a) shows the out-of-plane XRD profiles of the samples for Ts = 300ºC 135 –600ºC. All the results indicate (001)-oriented growth, and the appearance of 002 superlattice 136 peaks indicates the presence of B2 ordering. Figure 1(b) shows the corresponding XRD profiles 137 along the [111] direction at χ = 54.7°. The appearance of 111 superlattice peaks indicates the 138 presence of L21 ordering. Figure 1(c) shows the Ts dependence of 𝑆𝐵2 and 𝑆𝐿21. The 𝑆𝐵2 value 139 was 0.8 already at Ts = 300ºC, which further enhances to 0.9 at Ts = 600ºC. The 𝑆𝐿21  value 140 monotonically increases with Ts and reaches 0.8 at Ts = 600ºC. The first-principles 141 calculations  [32,36] reported that L21-ordered and B2-ordered MVG have negative spin 142 polarization, which is higher for the L21-ordered structure. The XRD results indicated that the 143 experimental samples had relatively high ordering, from which high negative spin polarization 144 can be expected.  145  146   We also examined the MVG growth on bottom electrodes using the series-B samples as a 147 preliminary step to fabricate CPP-GMR stacks. The structures of the bottom electrodes were Cr 148 (5)/Ag (50), Cr (5)/Ag (50)/V(5), and Cr (5)/Ag (50)/Cr (5)/W (15). We first used Cr/Ag bottom 149 electrodes to fabricate CPP-GMR stack because Cr/Ag is the standard bottom electrode employed 150 in studies on Heusler alloy-based CPP-GMR devices  [12]. However, we found that the 151 interdiffusion occurred between MVG and Ag (the results are shown in the supplementary 152 7  material). Thus, we searched for a suitable bottom electrode structure. The diffusion problem 153 between Mn and Ag is consistent with the report that Co2MnZ-type Heulser alloys showed lower 154 thermal stability with a Ag spacer than Co2FeZ-type Heulser alloys  [10]. In addition, the phase 155 diagram of Ag-Mn system shows a certain amount of Mn is miscible to Ag at the annealing 156 temperatures. Figures 1(d) and 1(e) show the out-of-plane and [111]-direction XRD profiles of 157 the series-B samples for Ts = 500ºC, respectively. The XRD peaks indicate the (001)-oriented 158 epitaxial growth of both bottom electrode and MVG layer in all the samples, regardless of the 159 various lattice misfits with respect to the MVG lattice constant reported for bulk study: MgO 160 (+0.9%), Ag (–2.2%), V (+2.6%), and W (+7%). Similar to the series-A samples, 002 and 111 161 superlattice peaks were observed indicating the presence of B2 and L21 ordering. In the case of 162 the Cr/Ag/Cr/W bottom electrode, the 𝑆𝐵2 and 𝑆𝐿21 values were calculated to be 0.99 and 0.82, 163 respectively, which were higher than those of the series-A sample at Ts = 500ºC. In this calculation 164 the XRD results of the sample only with the Cr/Ag/Cr/W bottom electrode shown as the dashed 165 line in Figs 1(d) and 1(e) were subtracted as a background. The order parameter of the sample 166 with Cr/Ag and Cr/Ag/V bottom electrodes were not analyzed because the MVA layer on these 167 bottom electrodes were not properly fabricated as discussed below by considering Ms of the MVG 168 layers. 169  170  171 8   172 Fig. 1. (a) Out-of-plane and (b) [111]-direction XRD profiles for the series-A samples for Ts = 173 300ºC –600ºC. The legends are common for (a) and (b), and the data are offset for clarity. (c) The 174 Ts dependence of SB2 and SL21 calculated from the peak intensities in (a) and (b). (d) Out-of-plane 175 and (e) [111]-direction XRD profiles for the series-B samples with Cr/Ag, Cr/Ag/V, and 176 Cr/Ag/Cr/W bottom electrodes for Ts = 500ºC. The data for the sample only with the Cr/Ag/Cr/W 177 bottom electrode and without the MVG layer is also shown. The legends are common for (d) and 178 (e), and the data are offset for clarity. 179  180 9   181   We next evaluated the magnetic properties of MVG. It has been reported that the saturation 182 magnetization of MVG is sensitive to disorder because of the ferrimagnetic configuration of the 183 Mn and V moments and thus can be a good indicator of the ordering  [36]. Figure 2(a) shows the 184 in-plane M-H curves of the series-A samples for Ts = 300ºC –600ºC. The M-H curves were similar 185 for Ts = 300ºC and 400ºC with Ms of 120 emu/cc and a coercive field (Hc) of approximately 800 186 Oe. By further increasing Ts, the Ms increased to 190 emu/cc at Ts = 500ºC and 220 emu/cc at Ts 187 = 600ºC, and the Hc decreased to 400 Oe at Ts = 600ºC. The Ms value is similar to those reported 188 in the previous study of MVG thin films  [36]. The Ms value estimated from the Slater-Pauling 189 rule based on the XRF composition was approximately 300 emu/cc. The experimental values 190 were smaller than the Slater-Pauling value reflecting the imperfect ordering. Although both Ms 191 and order parameters showed an increasing trend with respect to Ts, the trend is not completely 192 consistent. In the XRD results, 𝑆𝐵2 and 𝑆𝐿21 were almost the same between Ts = 400 and 500ºC, 193 while Ms changed significantly between these temperatures. Although the reason for the 194 inconsistency is unclear, this result suggests the importance of measuring both order parameters 195 and Ms to evaluate the quality of the MVG films. 196  197   Figure 2(b) shows the Ts dependence of Ms of the series-A samples calculated from the results 198 in Fig. 2(a). In addition, the Ms of the sample deposited at Ts = 300ºC and post-annealed at Tp = 199 500ºC is shown as an open square. The dashed line shows the Ms calculated from the Slater-200 Pauling rule. The Ms value for Ts = 300ºC + Tp = 500ºC was higher than those for Ts = 300 and 201 400ºC but lower than that for Ts = 500ºC, indicating that post-annealing at 500ºC improved the 202 ordering, though it was not as effective as Ts of the same temperature. This additional post-203 annealing will be employed later in the series-C sample. Figure 2(b) also shows the results of the 204 series-B samples. The Ms values for the three bottom electrode structures were measured at Ts = 205 500ºC. The Ms values of the Cr/Ag and Cr/Ag/V samples were significantly smaller than that of 206 10  the series-A sample, indicating that these bottom electrode structures have a detrimental effect on 207 the MVG growth. The reason for the small Ms of the Cr/Ag sample might be the interdiffusion 208 between MVG and Ag, as discussed above. The reasons for the small Ms of the Cr/Ag/V sample 209 might be interdiffusion between MVG and V, which disturbed the composition of MVG. In 210 contrast, the Ms of the Cr/Ag/Cr/W sample was slightly higher than the series-A samples. The Ts 211 dependence was measured for the Cr/Ag/Cr/W samples, which showed the similar trend to that 212 of the series-A samples, indicating that the W layer provided a good template for the growth on 213 MVG, despite the large misfit. Figure 2(c) shows the in-plane M-H curves of the Cr/Ag/Cr/W 214 samples for Ts = 300ºC–600ºC. In comparison with the series-A samples, the Hc was high 215 especially at lower Ts with Hc of ~2 kOe at Ts = 300 ºC. The saturation field (Hs) was also high 216 being ~7 kOe at Ts = 300 ºC. We speculate that the large Hc and Hs might be due to the dislocations 217 or lattice distortions at the interface between the MVG and W layers resulting from the large 218 misfit, which functioned as pinning sites. 219  220  221 11  Fig. 2. (a) M-H curves of the series-A samples for Ts = 300ºC –600ºC. (b) Dependence of Ms on 222 Ts for the series-A and series-B samples. The dashed line represents the value estimated from the 223 Slater-Pauling rule. (c) M-H curves of the series-B samples with the Cr/Ag/Cr/W bottom electrode 224 for Ts = 300–600ºC.  225  226   To demonstrate the negative spin polarization of MVG, we fabricated CPP-GMR stacks using 227 the Cr/Ag/Cr/W bottom electrode. The structure is shown in Fig. 3(a). Figures 3(b) and 3(c) show 228 the out-of-plane and [111]-direction XRD profiles of the series-C samples, respectively, for Ts = 229 300ºC–500ºC and Ts = 300ºC + Tp = 500ºC. 002 and 111 superlattice peaks were observed in all 230 the samples. Figure 3(d) shows 𝑆𝐵2 and 𝑆𝐿21 calculated from the data in Figs. 3(b) and 3(c). The 231 𝑆𝐵2 values indicate that the ordering was monotonically improved with increasing Ts and that 232 additional post-annealing can improve the ordering. This result is consistent with that of the 233 single-layer samples. The 𝑆𝐿21 values show a similar trend except that the value was higher at Ts 234 = 300ºC than Ts = 400ºC. A possible reason for this inconsistency is low accuracy because of the 235 weak peak intensities at Ts = 300ºC. Here, we did not evaluate Ms of the samples because the 236 thickness variations in the CoFe layer can cause large errors in the estimated Ms of the MVG layer 237 because CoFe has much higher Ms than MVG. We also examined the surface morphology of the 238 series-C samples. Figure 3(e) shows the Ra and P-V values with respect to the annealing conditions. 239 For Ts = 300ºC and 400ºC, the films were sufficiently flat to fabricate the CPP-GMR devices. 240 However, the flatness abruptly degraded at Ts = 500 ºC. The yield of the CPP-GMR devices from 241 this film was low due to the roughness, which will be discussed later. On the other hand, the 242 flatness for Ts = 300ºC + Tp = 500ºC was almost the same as that for Ts = 300ºC, indicating that 243 the additional post-annealing did not degrade the flatness.  244  245 12   246  247 Fig. 3. (a) Structure of the series-C samples. (b) Out-of-plane and (c) [111]-direction XRD profiles 248 of the series-C samples for Ts = 300ºC, 400ºC, 500ºC and Ts = 300ºC + Tp = 500ºC. The insets 249 show the enlarged view around the MVG 002 and 111 peaks. The legends are common for (b) 250 and (c), and the data are offset for clarity. (d) Annealing condition dependence of SB2 and SL21 251 calculated from the peak intensities in (b) and (c). (e) Annealing condition dependence of Ra and 252 P-V. 253  254   We measured the R-H curves from the CPP-GMR devices fabricated from the series-C samples. 255 Figures 4(a)–4(d) show examples of the R-H curves. The shapes of the R-H curves had large 256 13  device-to-device variations within each sample, which can be attributed to the following reason. 257 As shown in the M-H curves in Fig. 2(c), the magnetization of the MVG layer changed gradually 258 with H with low squareness, indicating that the field response of the MVG magnetization was 259 spatially non-uniform. Considering that the R-H curves reflected the local magnetic behavior of 260 the small device area, we speculate that the distributions in the field response resulted in variations 261 in the R-H curves. In addition, spatially non-uniform ordering of the MVG layer might lead to a 262 distribution in the MR ratio because the spin polarization is affected by the ordering. The R-H 263 curve in Fig. 4(a) for the upward H sweep can be interpreted as the schematics of the 264 magnetization configurations shown above the figure. The abrupt R change around the zero field 265 corresponds to the magnetization switching of the CoFe layer, and the slope of the R change 266 switches from negative to positive at this point. This interpretation indicates negative MR in 267 which R was higher (lower) when the magnetization configuration was more parallel (antiparallel). 268 Note that plateau regions with lower R corresponding to a complete antiparallel configuration 269 were not obtained, and the full MR ratio could not be evaluated. The negative MR provides 270 evidence of negative spin polarization in the MVG layer because both bulk spin asymmetry of 271 CoFe and interfacial spin scattering asymmetry of CoFe/Ag are positive  [42]. The bulk and 272 interfacial contributions to the negative spin polarization in the MVG layer cannot be separated 273 in this experiment, and further study is necessary to understand these properties. In the previous 274 TMR study on MVG reported similar R-H curves with an incomplete antiparallel magnetization 275 configuration due to gradual change in the magnetization in the MVG layer  [32]. 276   The R-H curve in Fig. 4(a) exhibits hysteresis in the large H range and the downward and upward 277 H sweeps coincide at around H = ±18kOe. Considering that the CoFe layer had small Hc, the 278 shape of the R-H curve is determined by the magnetization behavior of the MVG layer, and the 279 result indicated that the Hs of MVG layer was 18 kOe or higher, where the parallel configuration 280 was obtained. This large Hs is, to some extent, consistent with the M-H curve of the series-B 281 14  sample (Fig. 2(c)) in which the Hs was as high as 7 kOe at Ts = 300ºC. The difference in Hs 282 between the M-H and R-H curves might be attributed to the different thickness of the MVG layer 283 (37 nm in the series-B and 10 nm in the series-C sample). The large Hc and Hs were related to the 284 interface with the W layer presumably due to the dislocations or lattice distortion originating from 285 the large misfit between MVG and W. Thus, we speculate that the thinner MVG layer was more 286 affected by the interface with the W layer, resulting in higher Hs. Additionally, patterning of the 287 devices into nanopillars could affect the magnetization behavior due to pinning and different 288 domain formation  [43]. The inset shows the R-H curve in the H range used in the auto-prober 289 measurements. Because the R-H was measured only in the minor loop, the curve was asymmetric, 290 and the amplitude of MR ratio was smaller than that measured in the higher field range. Note that 291 the statistical data of the MR ratio presented in the following paragraph were obtained from the 292 minor loop. The R continued to change even at H = ±30 kOe, suggesting the existence of other 293 small contributions to R. 294  295   In the R-H curve at Ts = 400ºC Ts = 500ºC and Ts = 300ºC + Tp = 500ºC shown in Figs. 4(b), 4(c), 296 and 4(d), the downward and upward H sweeps coincide within the measurement field range, 297 because higher Ts or post-annealing reduced the Hs of the MVG layer. In addition, the amplitude 298 of the negative MR ratio increased. Figure 4(e) shows the summary of MR ratio obtained by auto-299 prober measurements of all the devices in each sample. The defective devices showing abnormal 300 R were excluded. The distributions in R and MR ratio are presented in the supplementary material. 301 The number of data points was significantly small in the case of Ts = 500ºC because of the low 302 device yield, which might be due to the increased roughness. The MR ratio shows large 303 distributions within each sample because of the shape variations in the R-H curves. When we 304 excluded the Ts = 500ºC sample because of the lack of statistical data, the MR ratio increased in 305 the following order: Ts = 300ºC, Ts = 400ºC, and Ts = 300ºC + Tp = 500ºC with the maximum 306 negative MR ratio being ~−1.8 %. The resistance area product (RA) of the Ts = 300ºC + Tp = 500ºC 307 15  sample was calculated to be 22 mΩ·μm2. Figure 4(f) shows the measurement temperature 308 dependence of the MR ratio for the Ts = 300ºC + Tp = 500ºC sample. The negative MR ratio was 309 enhanced by lowering the measurement temperature and reached ~−3.0%. 310  311  312 Fig. 4. (a)–(d) Examples of the R-H curves measured in the CPP-GMR devices fabricated from 313 the series-C samples for Ts = 300ºC, 400ºC, 500ºC and Ts = 300ºC + Tp = 500ºC. The solid and 314 dashed lines represent the downward and upward H sweeps, respectively. The schematics above 315 (a) depict the magnetization configurations of the MVG and CoFe layers for the upward H sweep, 316 and the inset in (a) shows the minor loop measured in the smaller H range. (e) Annealing condition 317 dependence of MR ratio of all the devices. (f) Measurement temperature dependence of MR ratio 318 for Ts = 300ºC + Tp = 500ºC. 319  320 16   321   To confirm that the negative MR was obtained in the GMR stack properly fabricated as designed, 322 we observed the cross-sectional microstructure of the series-C sample for Ts = 300ºC + Tp = 500ºC. 323 Figures 5(a) and 5(b) show a HAADF-STEM image and EDS mapping, respectively. The results 324 confirm sharp and flat interfaces between the layers without noticeable interdiffusion. The 325 thicknesses of the layers were close to the designed values. Figure 5(c) shows the EDS line profile 326 along the growth direction. The MVG layer composition was estimated to be Mn54.8V20.5Ga24.8, 327 which was in good agreement with the XRF results. The CoFe layer composition was close to 328 Co50Fe50, as designed. Figure 5(d) shows the NBED patterns obtained in the MVG layer. The 002 329 and 111 superlattice spots were observed, confirming the existence of the B2 and L21 ordering. 330 Figure 5(e) shows the atomic-resolution HAADF-STEM image and the corresponding Fourier-331 transform filtered image obtained at the upper interface of MVG with Ag. An atomically flat 332 matched interface was observed. The interface has a few dislocations marked by ⊥ symbols. 333 Figure 5(f) shows the similar results at the lower interface of MVG with W. The lower interface 334 was rougher than the upper interface, indicating that the flatness was improved in the MVG layer 335 by appropriate thermal treatment.  336  337 17   338 Fig. 5. (a) HAADF-STEM image and (b) EDS mapping of the series-C sample for Ts = 300ºC + 339 Tp = 500ºC. (c) EDS line profile obtained from (b). (d) NBED patterns obtained in the MVG layer. 340 (e) Atomic-resolution HAADF-STEM image and the corresponding Fourier-transform filtered 341 image obtained at (e) the upper interface of MVG with Ag and (f) the lower interface of MVG 342 with W. The yellow ⊥ symbols indicate dislocations. 343  344   Finally, we calculated the transmittance of MVG/Ag/MVG(001) to examine the Fermi surface 345 matching of MVG with Ag spacer for utilizing the negative spin polarization of MVG. Figures 346 6(a) and 6(b) show in-plane wave vector dependence of the minority-spin transmittance in the 347 parallel magnetization configuration for Mn and VGa termination, respectively. We confirmed 348 that the magnetic structure of MVG in MVG/Ag/MVG(001) is consistent with that of bulk L21-349 MVG obtained in the previous work35. The transmittance originates dominantly from the 350 minority-spin channel, reflecting the negative spin polarization of MVG. From the transmittance, 351 the interfacial RA was calculated to be 6.97 mΩ·μm2 for Mn termination and 5.86 mΩ·μm2 for 352 VGa termination. For comparison, we refer to the cases of Co2FeGa0.5Ge0.5 / Ag / 353 Co2FeGa0.5Ge0.5(001), in which a high MR ratio has been reported experimentally. The interfacial 354 18  RA of majority-spin electron in this systems was calculated to be 1.77, 1.69, and 2.32 mΩ·μm2 355 for FeGa, FeGe and, Co termination, respectively  [12,44] , which are much lower than that of 356 minority-spin electron in MVG / Ag / MVG(001). It has been reported that the calculated 357 interfacial RA and experimental MR ratio have negative correlation; the lower RA tends to have 358 higher MR ratio  [45]. Therefore, relatively high interfacial RA of MVG / Ag / MVG(001) 359 reflecting weak hybridization between Ag and Mn (V) orbitals suggests that the Ag spacer is not 360 suitable for MVG, which partly explains the small magnitude of negative MR ratio obtained in 361 this study. Further research is necessary to find an optimal spacer material.  362  363  364 Fig. 6. In-plane wave vector dependence of the minority-spin transmittance in parallel 365 magnetization configuration calculated for Mn2VGa / Ag / Mn2VGa(001) with (a) Mn termination 366 and (b) VGa termination. 367  368   This work demonstrated the negative MR ratio in the epitaxial MVG/Ag/CoFe pseudo-spin-369 valve CPP-GMR devices, providing evidence of the negative spin polarization in MVG. Despite 370 the large device-to-device distributions in the measured MR ratio within each sample, reflecting 371 the variations in the field response and ordering of the MVG layer in the small device area, the 372 MR ratio exhibited a statistical change with the annealing conditions and was enhanced in the 373 following order: Ts = 300ºC, Ts = 400ºC, and Ts = 300ºC + Tp = 500ºC. This trend coincides with 374 the trend of Ms for these thermal treatments, and the increase in the MR ratio is attributed to both 375 reduced Hs and improved ordering by optimizing the annealing conditions. The maximum MR 376 19  value at room temperature was ~−1.8 % for Ts = 300ºC + Tp = 500ºC. Note that this value is 377 underestimated because of the incomplete antiparallel configuration. Interestingly, the R-H curves 378 for Ts = 300ºC showed a large hysteresis, and the MVG layer acted as a fixed magnetic layer in 379 the minor loop measured in the small H range. This behavior is applicable to sensor operation, 380 though the MR ratio is small. According to the Ts dependence of Ms shown in Fig. 2(b), Ts = 381 500ºC and above significantly improved the ordering of the MVG grown on the MgO substrates 382 or W layers. However, such an annealing condition increased the roughness and was unsuitable 383 for the CPP-GMR stacks. In addition, the transmittance calculation indicated that the Ag spacer 384 is not suitable for MVG. In these aspects, the potential of MVG has not been fully exploited. To 385 this end, suitable template materials to facilitate ordering and improve the flatness of MVG, 386 adjusting MVG composition to promote ordering at lower temperatures, optimal thermal 387 treatments, and suitable spacer materials needs to be further studied. 388  389  390 IV CONCLUSIONS 391   We studied the negative spin polarization of MVG in epitaxial pseudo-spin-valve CPP-GMR 392 devices. We used a W insertion layer to prevent interdiffusion between the MVG layer and the 393 low-resistance Ag bottom electrode. The MVG layer on the W layer exhibited similar Ms to those 394 on MgO substrates, indicating that the W layer did not disturb the ordering of MVG. We 395 fabricated the GMR stack consisting of MVG/Ag/CoFe, and TEM observation confirmed the 396 sharp and flat interfaces between each layer without noticeable interdiffusion. Negative MR ratio 397 was observed in the CPP-GMR devices, and the MR value changed with thermal treatments 398 because the thermal treatments affected the ordering, roughness, and magnetic properties of MVG. 399 The maximum negative MR ratio was approximately –1.8% at room temperature. This value is 400 the highest negative MR reported in pseudo-spin-valve CPP-GMR devices, even though it is 401 20  underestimated due to the incomplete antiparallel magnetization configuration. These results 402 demonstrated the potential of MVG as a material with high negative spin polarization. 403  404 SUPPLEMENTARY MATERIAL 405   The supplementary material explains TEM and XRD data of CPP-GMR stack using Cr/Ag 406 bottom electrodes and distributions of R and MR in all the devices fabricated on series-C sample 407 for Ts = 300ºC + Tp = 500ºC. 408  409 ACKNOWLEDGMENTS  410 This work was partially supported by Advanced Storage Research Consortium (ASRC), JSPS 411 KAKENHI Grant No. 21K20434, and No. 23K03934, and MEXT Initiative to Establish Next-412 generation Novel Integrated Circuits Centers (X-nics) Grant No. JPJ011438. The authors thank 413 Dr. T. Kubota of Tohoku University for fruitful discussions and M. Inoue of NIMS for technical 414 support. The calculations in this study were performed on the Numerical Materials Simulator at 415 NIMS. 416  417 CONFLICT OF INTEREST STATEMENT 418 The authors declare no conflicts of interest associated with this manuscript. 419  420 DATA AVAILABILITY SATATEMENT 421 The data that support the findings of this study are available from the corresponding author upon 422 reasonable request.  423  424   425 21  References 426  427 [1] J. C. Slonczewski, Current-Driven Excitation of Magnetic Multilayers, J. Magn. Magn. 428 Mater. 159, L1 (1996). 429 [2] L. Berger, Emission of Spin Waves by a Magnetic Multilayer Traversed by a Current, 430 Phys. Rev. B 54, 9353 (1996). 431 [3] S. I. Kiselev, J. C. Sankey, I. N. Krivorotov, N. C. Emley, R. J. Schoelkopf, R. A. 432 Buhrman, and D. C. Ralph, Microwave Oscillations of a Nanomagnet Driven by a Spin-433 Polarized Current, Nature 425, 380 (2003). 434 [4] M. Hosomi et al., A Novel Nonvolatile Memory with Spin Torque Transfer 435 Magnetization Switching: Spin-RAM, Tech. Dig. - Int. Electron Devices Meet. IEDM 436 2005, 459 (2005). 437 [5] F. Macià, A. D. Kent, and F. C. Hoppensteadt, Spin-Wave Interference Patterns Created 438 by Spin-Torque Nano-Oscillators for Memory And, Nanotechnology 22, 095301 (2011). 439 [6] J. Torrejon et al., Neuromorphic Computing with Nanoscale Spintronic Oscillators, Nat. 440 2017 5477664 547, 428 (2017). 441 [7] T. Kanao, H. Suto, K. Mizushima, H. Goto, T. Tanamoto, and T. Nagasawa, Reservoir 442 Computing on Spin-Torque Oscillator Array, Phys. Rev. Appl. 12, (2019). 443 [8] Y. Sakuraba, M. Hattori, M. Oogane, Y. Ando, H. Kato, A. Sakuma, T. Miyazaki, and 444 H. Kubota, Giant Tunneling Magnetoresistance in Co2MnSi/Al-O /Co 2MnSi Magnetic 445 Tunnel Junctions, Appl. Phys. Lett. 88, 10 (2006). 446 [9] H. Liu, Y. Honda, T. Taira, K. Matsuda, M. Arita, T. Uemura, and M. Yamamoto, Giant 447 Tunneling Magnetoresistance in Epitaxial Co2MnSi/MgO/Co2MnSi Magnetic Tunnel 448 Junctions by Half-Metallicity of Co2MnSi and Coherent Tunneling, Appl. Phys. Lett. 449 101, 132418 (2012). 450 22  [10] S. Li, Y. K. Takahashi, T. Furubayashi, and K. Hono, Enhancement of Giant 451 Magnetoresistance by L21 Ordering in Co2Fe(Ge0.5Ga0.5) Heusler Alloy Current-452 Perpendicular-to-Plane Pseudo Spin Valves, Appl. Phys. Lett. 103, 042405 (2013). 453 [11] M. Jourdan et al., Direct Observation of Half-Metallicity in the Heusler Compound 454 Co2MnSi, Nat. Commun. 5, 3974 (2014). 455 [12] J. W. Jung, Y. Sakuraba, T. T. Sasaki, Y. Miura, and K. Hono, Enhancement of 456 Magnetoresistance by Inserting Thin NiAl Layers at the Interfaces in 457 Co2FeGa0.5Ge0.5/Ag/Co2FeGa0.5Ge0.5 Current-Perpendicular-to-Plane Pseudo Spin 458 Valves, Appl. Phys. Lett. 108, 102408 (2016). 459 [13] S. Andrieu, A. Neggache, T. Hauet, T. Devolder, A. Hallal, M. Chshiev, A. M. Bataille, 460 P. Le Fèvre, and F. Bertran, Direct Evidence for Minority Spin Gap in the <math> 461 <mrow> <mi Mathvariant="normal">C</Mi> <msub> <mi 462 Mathvariant="normal">o</Mi> <mn>2</Mn> </Msub> <mi>MnSi</Mi> </Mrow> 463 </Math> Heusler Compound, Phys. Rev. B 93, 094417 (2016). 464 [14] K. Elphick, W. Frost, M. Samiepour, T. Kubota, K. Takanashi, H. Sukegawa, S. Mitani, 465 and A. Hirohata, Heusler Alloys for Spintronic Devices: Review on Recent Development 466 and Future Perspectives, Sci. Technol. Adv. Mater. 22, 235 (2021). 467 [15] K. Hamaya and M. Yamada, Semiconductor Spintronics with Co2-Heusler Compounds, 468 MRS Bull. 47, 584 (2022). 469 [16] M. AlHajDarwish, H. Kurt, S. Urazhdin, A. Fert, R. Loloee, W. P. Pratt, and J. Bass, 470 Controlled Normal and Inverse Current-Induced Magnetization Switching and 471 Magnetoresistance in Magnetic Nanopillars, Phys. Rev. Lett. 93, 157203 (2004). 472 [17] H. Suto, T. Nakatani, N. Asam, H. Iwasaki, and Y. Sakuraba, Evaluation of Spin-473 Transfer-Torque Efficiency Using Magnetization Reversal against a Magnetic Field: 474 Comparison of FeCr with Negative Spin Polarization and NiFe, Appl. Phys. Express 16, 475 013003 (2023). 476 23  [18] J.-G. Zhu, X. Zhu, and Y. Tang, Microwave Assisted Magnetic Recording, IEEE Trans. 477 Magn. 44, 125 (2008). 478 [19] S. Bosu, H. Sepehri-Amin, Y. Sakuraba, S. Kasai, M. Hayashi, and K. Hono, High 479 Frequency Out-of-Plane Oscillation with Large Cone Angle in Mag-Flip Spin Torque 480 Oscillators for Microwave Assisted Magnetic Recording, Appl. Phys. Lett. 110, 142403 481 (2017). 482 [20] W. Zhou, H. Sepehri-Amin, T. Taniguchi, S. Tamaru, Y. Sakuraba, S. Kasai, H. Kubota, 483 and K. Hono, Inducing Out-of-Plane Precession of Magnetization for Microwave-484 Assisted Magnetic Recording with an Oscillating Polarizer in a Spin-Torque Oscillator, 485 Appl. Phys. Lett. 114, 172403 (2019). 486 [21] N. Asam, H. Suto, S. Tamaru, H. Sepehri-Amin, A. Bolyachkin, T. Nakatani, W. Zhou, 487 H. Kubota, and Y. Sakuraba, Analysis Method of a Spin-Torque Oscillator Using Dc 488 Resistance Change during Injection Locking to an External Microwave Magnetic Field, 489 Appl. Phys. Lett. 119, 142405 (2021). 490 [22] M. Takagishi, N. Narita, H. Iwasaki, H. Suto, T. Maeda, and A. Takeo, Design Concept 491 of MAS Effect Dominant MAMR Head and Numerical Study, IEEE Trans. Magn. 57, 492 3300106 (2021). 493 [23] Y. Nakagawa, M. Takagishi, N. Narita, A. Takeo, and T. Maeda, Multiple Spin Injection 494 Into Coupled Field Generation Layers for Low Current Operation of MAMR Heads, 495 IEEE Trans. Magn. 58, 3201005 (2022). 496 [24] K. Sunaga, M. Tsunoda, K. Komagaki, Y. Uehara, and M. Takahashi, Inverse Tunnel 497 Magnetoresistance in Magnetic Tunnel Junctions with an Fe4N Electrode, J. Appl. Phys. 498 102, 013917 (2007). 499 [25] S. Yasui, S. Honda, J. Okabayashi, T. Yanase, T. Shimada, and T. Nagahama, Large 500 Inverse Tunnel Magnetoresistance in Magnetic Tunnel Junctions with an <math 501 Display="inline" Overflow="scroll"> <msub> <mi>Fe</Mi> <mn>3</Mn> 502 24  </Msub> <msub> <mrow> <mrow> <mi Mathvariant="normal">O</Mi> </Mrow> 503 </Mrow> <mn>4</Mn> </Msub> </Math> Ele, Phys. Rev. Appl. 15, 034042 (2021). 504 [26] C. Bellouard, H. Rapp, B. George, S. Mangin, G. Marchal, and J. Ousset, Negative Spin-505 Valve Effect i/A Trilayers, Phys. Rev. B - Condens. Matter Mater. Phys. 53, 5082 (1996). 506 [27] C. Vouille, A. Barthélémy, F. E. Mpondo, A. Fert, P. A. Schroeder, S. Y. Hsu, A. Reilly, 507 and R. Loloee, Microscopic Mechanisms of Giant Magnetoresistance, Phys. Rev. B - 508 Condens. Matter Mater. Phys. 60, 6710 (1999). 509 [28] X. Li, H. Li, M. Jamali, and J.-P. Wang, Damping Constant Measurement and Inverse 510 Giant Magnetoresistance in Spintronic Devices with Fe4N, AIP Adv. 7, 125303 (2017). 511 [29] H. Suto, T. Nakatani, Y. Kota, A. Nagarjuna, H. Iwasaki, K. Amemiya, T. Mitsui, S. 512 Sakai, S. Li, and Y. Sakuraba, Study on FeCr Thin Film for a Spintronic Material with 513 Negative Spin Polarization, J. Magn. Magn. Mater. 557, 169474 (2022). 514 [30] K. Ramesh Kumar, N. Harish Kumar, G. Markandeyulu, J. A. Chelvane, V. Neu, and P. 515 D. Babu, Structural, Magnetic and Transport Properties of Half-Metallic Ferrimagnet 516 Mn2VGa, J. Magn. Magn. Mater. 320, 2737 (2008). 517 [31] K. Ramesh Kumar, N. Harish Kumar, P. D. Babu, S. Venkatesh, and S. Ramakrishnan, 518 Investigation of Atomic Anti-Site Disorder and Ferrimagnetic Order in the Half-Metallic 519 Heusler Alloy Mn 2 V Ga, J. Phys. Condens. Matter 24, 336007 (2012). 520 [32] C. Klewe, M. Meinert, J. Schmalhorst, and G. Reiss, Negative Spin Polarization of 521 Mn2VGa Probed by Tunnel Magnetoresistance, J. Phys. Condens. Matter 25, 76001 522 (2013). 523 [33] K. Ozdogan, I. Galanakis, E. Şaşioglu, and B. Aktaş, Search for Half-Metallic 524 Ferrimagnetism in V-Based Heusler Alloys Mn 2VZ (Z ≤ Al, Ga, In, Si, Ge, Sn), J. Phys. 525 Condens. Matter 18, 2905 (2006). 526 [34] L. Wollmann, S. Chadov, J. Kübler, and C. Felser, Magnetism in Cubic Manganese-Rich 527 Heusler Compounds, Phys. Rev. B 90, 214420 (2014). 528 25  [35] L. Fan, F. Chen, C. Li, X. Hou, X. Zhu, J. Luo, and Z.-Q. Chen, Promising Spintronics: 529 Mn-Based Heusler Alloys Mn3Ga, Mn2YGa (Y = V, Nb, Ta), ScMnVGa, J. Magn. Magn. 530 Mater. 497, 166060 (2020). 531 [36] Z. H. Li et al., Enhancing Atomic Ordering, Magnetic and Transport Properties of 532 Mn2VGa Heusler Alloy Thin Films toward Negatively Spin-Polarized Charge Injection, 533 Acta Mater. 276, 120110 (2024). 534 [37] N. Narita, M. Takagishi, H. Iwasaki, H. Suto, G. Koizumi, A. Takeo, and T. Maeda, 535 Design and Numerical Study of Flux Control Effect Dominant MAMR Head: FC Writer, 536 IEEE Trans. Magn. 57, 3300205 (2021). 537 [38] R. Landauer, Electrical Resistance of Disordered One-Dimensional Lattices, Philos. 538 Mag. 21, 863 (1970). 539 [39] P. Giannozzi et al., QUANTUM ESPRESSO: A Modular and Open-Source Software 540 Project for Quantum Simulations of Materials, J. Phys. Condens. Matter 21, 395502 541 (2009). 542 [40] P. Giannozzi et al., Advanced Capabilities for Materials Modelling with Quantum 543 ESPRESSO, J. Phys. Condens. Matter 29, 465901 (2017). 544 [41] J. P. Perdew, K. Burke, and M. Ernzerhof, Generalized Gradient Approximation Made 545 Simple, Phys. Rev. Lett. 77, 3865 (1996). 546 [42] J. W. Jung, Z. Jin, Y. Shiokawa, and M. Sahashi, Investigations of Interface Spin 547 Asymmetry and Interfacial Resistance in Fe x Co100 − x /Ag Interface, J. Appl. Phys. 548 117, 17A323 (2015). 549 [43] T. Seki, H. Iwama, T. Shima, and K. Takanashi, Size Dependence of the Magnetization 550 Reversal Process in Microfabricated L1 0 -FePt Nano Dots, J. Phys. D. Appl. Phys. 44, 551 335001 (2011). 552 26  [44] K. Simalaotao, Y. Miura, and Y. Sakuraba, Origin of the Difference in Spin-Dependent 553 Transport Properties between Half-Metallic Co2FeGa0.5Ge0.5 and Co2Mn0.5Fe0.5Si 554 with the Ag Spacer, Prep. (n.d.). 555 [45] Y. Miura, First-Principles Studies on Spin-Dependent Transport of Magnetic Junctions 556 with Half-Metallic Heusler Compounds, J. Magn. Soc. Japan 47, 2307R005 (2023). 557  558   559 27  Supplementary material 560 Negative spin polarization of Mn2VGa Heusler alloy thin films studied in current-561 perpendicular-to-plane giant magnetoresistance devices 562 Hirofumi Suto, Vineet Barwal, Kodchakorn Simalaotao, Zehao Li, Keisuke Masuda, Taisuke 563 Sasaki, Yoshio Miura, and Yuya Sakuraba 564 Research Center for Magnetic and Spintronic Materials, National Institute for Materials Science 565 (NIMS), Tsukuba, 305-0047, Japan 566  567 STEM observation and XRD of CPP-GMR stack containing a Cr/Ag bottom electrode 568   Figures S1(a) show the structure of the CPP-GMR stack containing a Cr/Ag bottom electrode 569 and a 20-nm-thick MVG layer deposited at 550ºC. Figures S1(b) and S1(c) show a HAADF-570 STEM image and EDS mapping, respectively. The TEM observation shows the formation of Ag 571 columns penetrating the MVG layer, disrupting the designed layer structure. Because The layers 572 above the MVG layer were deposited at room temperature after cooling down, their structure was 573 not damaged. The electrical current is expected to flow preferentially through the Ag columns, 574 and this stack is not suitable to evaluate the spin polarization of MVG. Figure S1(d) shows the 575 EDS line profile obtained from the region where MVG layer exist. A small amount of Mn and Ga 576 was detected in the whole Ag bottom electrode. This diffusion of Mn and Ga into the Ag bottom 577 electrode is consistent with the phase diagram of Ag-Mn and Ag-Ga system, which shows small 578 amount of Mn and Ga is miscible to Ag at the annealing temperature. Figure S1(e) shows the EDS 579 line profile obtained from the region with the Ag column. The composition of the column though 580 the MVG layer is almost the same as that of the bottom electrode.  581   Figures S2(a) and S2(b) show the out-of-plane and [111]-direction XRD profiles. Despite the 582 disrupted layer structure, (001)-oriented growth of the MVG layer was maintained, and 002 and 583 111 super lattice peaks were observed.  584  585 28   586 FIG. S1. (a) Structure of the CPP-GMR stack containing a Cr/Ag bottom electrode and a MVG 587 layer. (b) HAADF-STEM image and (c) EDS mapping. EDS line profile obtained from (d) the 588 region where MVG layer exist and (e) the region with Ag column through the MVG layer. The 589 dotted lines in (c) shows the position of the EDS line profiles. 590  591  592  593 FIG. S2. (a) Out-of-plane and (b) [111]-direction XRD profiles of the CPP-GMR stack shown in 594 Fig. S1. 595   596 29   597 Resistance and MR ratio distribution of CPP-GMR devices for Ts = 300ºC + Tp = 500ºC 598   Figures S3 show the Rmax and MR ratio of the 75 devices with dimensions of (a) 80 nm × 80 nm 599 and (b) 140 nm × 70 nm fabricated from the series-C samples for Ts = 300ºC + Tp = 500ºC, 600 respectively. The data from 5 devices showing abnormal R were excluded from the analysis. 601  602  603 FIG. S3. Rmax and MR ratio of the devices fabricated the series-C samples for Ts = 300ºC + Tp = 604 500ºC. The device size is 80 nm × 80 nm in (a) and 140 nm × 70 nm in (b). 605