VdW heterostructure interlayer spacing regulation via laser shocking
Figure 1a shows a typical MRAM composed of a matrix of spin valves consisting of vdW heterostructures based on EB effect. To obtain higher performance requirements such as high tunneling magnetoresistance and high read fault tolerance5. 7, the modulation of the EB effect can enhance the interlayer coupling to optimize the performance. According to the Mauri model35, Heb is inversely proportional to the interlayer spacing. Therefore, the enhancement of the EB field and saturation magnetization can be achieved by adjusting the interlayer spacing (Fig. 1b). Here, FGT and FPSe single crystals were confirmed by scanning electron microscope and energy dispersive spectrometer (see Supplementary Figs. 1 and 2), and their bulk magnetizing characteristics are shown in Supplementary Fig. 3, indicating a Néel temperature TN of 113 K for FPSe and a Curie temperature TC of 230 K for FGT, consistent with those reported in the literature36–39. To study the interlayer coupling tuned by vdW spacing engineering, we prepared bilayer heterostructures composed of FGT and FPSe flakes. Raman spectrum was used to characterize FPSe, FGT, and FPSe/FGT heterostructure, where the peaks could well match those of FPSe and FGT, indicating excellent quality of the FPSe/FGT heterostructure (Supplementary Fig. 4). The LS technology enables the modulation of the interlayer spacing of the vdW heterostructures as a technical method, which can provide high pressure up to ∼20 GPa peak level in an ultrashort time scale (tens of picosecond), like a “hammering” operation. As shown in Fig. 1c, the vdW heterostructure consists of FPSe/FGT with a natural vdW interlayer distance of d1, corresponding to a weak interlayer coupling. After LS40, 41, its vdW interlayer distance will be reduced to d2, promising a strong interlayer coupling. The evolution of interlayer spacing under LS was studied employing molecular dynamics (MD) simulations (Fig. 1d). The equilibrium states were simulated with vdW interlayer spacing at both the unstrained configuration and strained configuration. This simulation shows that the FPSe/FGT system has two different equilibrium configurations. The first one presents an inter-flake spacing (distance between flake edge atoms) of 6.6 Å, while the second one emerges when the inter-flake distance is 4.2 Å (see Supplementary Information for calculation details, Section 3 and Section 4). Supplementary Fig. 5–8 and Table 1–3 show the calculation details and strain evolution in the LS MD simulation. More importantly, LS is effective to flatten the wrinkles and voids without any damage to the heterostructures42, 43. Further, we used cross-sectional STEM image to measure the vdW gap width of FPSe/FGT heterostructure. Performed with LS, vdW interlayer distances of FPSe and FGT were ~ 2.5 Å and ~ 2.1 Å respectively, which are basically consistent with previous reports39, 44–46, but the original gap width of vdW heterostructures (~ 7.4 Å) was dramatically reduced to ~ 4.6 Å (Fig. 1e, f). The simulation results verify the reduction of the interlayer distance upon pressure, which is consistent with the experimental results.
Enhanced interfacial magnetic coupling by interlayer spacing modulation
To confirm the enhancement of HEB and Tb in AFM/FM heterosturcture via LS technology, magneto-optical Kerr effect (MOKE) techniques were used to investigate the magnetic properties of FPSe/FGT heterostructure, denoted as sample A (Supplementary Fig. 9a). Sample A was characterized by atomic force microscope, while the thicknesses of FPSe was ~ 24.4 nm and that of FGT was ~ 18.0 nm (see Supplementary Fig. 9b). We first explored the relationship of Kerr rotation angle (θk) with respect to the magnetic field of FPSe/FGT before LS. Figure 2b shows the typical temperature-dependent Kerr loops of FPSe/FGT as a function of magnetic field (B) from 5 K to 80 K, where θk vs B exhibits an obvious shift along the negative B direction, signifying the emergence of HEB. However, for FGT flake before LS, the typical MOKE curves of FGT as a function of magnetic field (B) from 5 K to 200 K were measured (Supplementary Fig. 10), and their Kerr hysteresis loops remained symmetric relative to the zero point along the B-axis, suggesting a lack of HEB. The EB effect could be explained by the Meiklejohn-Bean model (namely M-B model, FPSe and FGT both with a single domain state). In the AFM/FM system, an adjacent AFM layer causes a unidirectional pinning of the FM, manifesting as a shift in the hysteresis loop along the magnetic field axis below the Néel temperature (TN) of the AFM, known as the EB field HEB (Supplementary Fig. 11). With the increase of temperature, HEB decreases gradually and disappears above 20 K, indicating a Tb of EB field at 20 K.
On the other hand, according to the Mauri model, HEB can be written as:
where A12 is the interfacial exchange stiffness, ξ is the distance of the interlayer, MF is the saturation magnetization of the ferromagnet and tF is the FM thickness, suggesting that HEB is negatively correlated with interlayer distance ξ and FM layer thickness tF. Eq. (1) reveals that vdW spacing engineering can manipulate the magnetic coupling in vdW FPSe/FGT heterostructure. Therefore, we utilized LS to perform vdW spacing engineering, inducing extreme out-of-plane pressure onto another vdW FPSe/FGT heterostructure. Supplementary Table 4 provides a relationship of laser power and pressure in Supplementary Information Section 5. Its EB effect was effectively enhanced, as shown in Fig. 2c. The EB effect and Tb of sample A without/with LS are shown in Fig. 2d and 2e summarized in Fig. 2f. After LS, HEB increased from 29.1 mT to 111.2 mT at 5 K, and Tb increased from 20 K to 110 K, which is close to the Néel temperature of FPSe (113 K). The prominent improvement of HEB and Tb confirms that the interlayer coupling of 2D magnetic heterostructures can be strongly enhanced by LS. Furthermore, Fig. 2g and 2h demonstrate the significant enhancement of the coercive field (HC−L: left coercive field, HC−R: right coercive field), which can also be verified from Fig. 2i. In Fig. 2i, to further confirm the enhancement of the coercive field, an on-off ratio parameter R is defined as \(\text{R}\text{ =}\frac{{\text{θ}}_{\text{ku(-180 mT)}}\text{ }\text{-}\text{ }{\text{θ}}_{\text{kd(-180 mT)}\text{ }}}{{\text{θ}}_{\text{ku(0 mT)}}\text{ }\text{-}\text{ }{\text{θ}}_{\text{kd(0 mT)}\text{ }}}\), where θku(−180 mT), θkd(−180 mT), θku(0 mT), and θkd(0 mT) represent the magneto-optical Kerr angles at − 180 mT and 0 mT, corresponding to up and down sweeping directions of the hysteresis loop, respectively. The significantly enhanced coercive field offers a remarkable advantage in obtaining a large field window to reduce the error rate of data writing and reading during spin valves and MTJs design.
Comparison between interlayer spacing modulation and thickness modulation
According to the Mauri model, HEB is not only modulated by the interlayer spacing, but can also be influenced by the material thickness (Fig. 3a). For the FPSe/FGT heterostructure as exchange-biased AFM/FM systems, the effect of FM or AFM thickness on HEB was investigated. The anisotropy energy that can pin the FM spins during magnetization reversal, is enhanced upon increasing the AFM thickness, and HEB is also strengthened concomitantly. Figure 3b shows the dependences of AFM thickness tFPSe on HEB, where HEB increases from 28 mT (FPSe (17.2 nm)/FGT (26.9 nm)) to 62 mT (FPSe (21.6 nm)/FGT (26.9 nm)). To further explore the effect of tFPSe on HEB, the FPSe/FGT heterostructures (FPSe with different thicknesses, and FGT with certain thickness of ~ 18.0 nm) were prepared, and the corresponding results are shown in Supplementary Fig. 12a. In addition, we also investigated the dependences of HEB on the thickness of FM (tFGT), and an inverse relationship between the EB effect and the thickness of FGT was observed. Figure 2f displays the hysteresis loops for FPSe (~ 27.0 nm)/FGT (18.0 nm, 22.4 nm, and 25.5 nm, respectively) heterostructures at 5 K, where HEB decreases from 32 mT (FPSe/FGT, 27.0 nm/18.0 nm) to 10 mT (FPSe/FGT, 27.0 nm/25.5 nm). The dependences of HEB on tFGT are summarized in Supplementary Fig. 12b, showing that HEB decreases upon increasing tFGT, which satisfies the FM-thickness dependence of HEB in conventional exchange-biased systems. In contrast, interlayer spacing modulation becomes a more stable and efficient modulation method with lower cost. According to the above experimental results, the weak EB effect in FPSe/FGT heterostructure is extremely sensitive to its interlayer spacing. We have investigated the effect of various out-of-plane pressures on the EB effect, and temperature-dependent Kerr loops of FPSe/FGT heterostructures (sample B, optical image in Supplementary Fig. 13) pressed under 0 GPa, ~ 8 GPa, ~ 11 GPa, and ~ 13 GPa at 5 K were measured respectively (Fig. 3d). Under 0 GPa, a slightly shifted loop signifies weak magnetic interface coupling in FPSe/FGT heterostructures. As the pressure increases, HEB is modulated to enhance. Tb is also sensitive to the pressure, increasing from 20 K to 110 K that is close to the Néel temperature, when the pressure varies from 0 GPa to ~ 13 GPa. The coercive field (HC) is also remarkably enhanced as the pressure increases, as shown in Supplementary Fig. 14. The performance comparison before and after LS is shown in Supplementary Table 5. Figure 3e-f illustrate the evolution of HEB in thickness-modulation and interlayer spacing-modulation. Furthermore, we compared HEB and Tb as a function of pressure and tAFM/tFM, as shown in Fig. 3g and 3h. HEB and Tb show a linear correlation with the increase of pressure by fitting. Compared to thickness-modulation, the interlayer spacing-modulation results reveal a clear pressure-induced magnetic coupling enhancement, offering important insights into the understanding of the physical nature of the interlayer magnetic states.
Based on the above experimental results, we fabricated high-quality tunneling spin valves (FPSe/FGT/h-BN/FGT, with optical image and atomic force microscope image illustrated in Supplementary Figs. 15 and 16) using the dry transfer technique in a glove box to investigate the spin valve behavior under various modulation methods. As shown in Fig. 3i, states 0–2 represent the original spin valve, the spin valve under thickness-modulation, and the spin valve under interlayer spacing-modulation, respectively. To demonstrate the occurrence of spin valve behavior, the resistance of the MTJ as a function of a perpendicular magnetic field was measured. Figure 3j shows the result of measurements performed at 5 K in the three states. For state 0 (optical image in Supplementary Fig. 15), as B was swept from negative to positive values, the resistance suddenly increased from approximately ~ 20 to ~ 49 kΩ at ~ 40 mT, followed by a sudden decrease back to ~ 20 kΩ at ~ 80 mT. As the magnetic field was swept back, an analogous abrupt increase and equally abrupt decrease in tunneling magnetoresistance were observed at ~ − 80 mT and − 140 mT. This was precisely the behavior expected for a tunneling spin-valve due to the hysteresis in the switching of the magnetization of the two ferromagnetic electrodes. For state 1 (optical image in Supplementary Fig. 16), the thicknesses of top FPSe and FGT were different from that of state 0. For state 2 (optical image in Supplementary Fig. 15), the resistance of the MTJ as a function of a perpendicular magnetic field was measured by using the spin valve device of state 0, under interlayer spacing-modulating. Similar spin valve behavior was observed in state 1 and state 2 (Fig. 3j). And Supplementary Fig. 17 shows the resistance of the magnetic tunnel junction as a function of magnetic field (B) from 5 K to 120 K. The magnitude of the TMR was defined as (RAP – RP)/RP, which characterizes the transmission efficiency, where RAP and RP represent the resistance obtained for parallel and antiparallel alignments of the magnetization. The TMR values were determined to be 141%, 130%, and 154% at 5 K for states 0–2, respectively. Due to the enhancement of the coercive field after modulation, the field window of state 0 increases from 60 mT to 140 mT (state 1) and 320 mT (state 2), respectively. Compared to thickness-modulation, interlayer spacing-modulation is more stable and efficient with a reduced cost. In addition, the TMR and the field window of the FGT/h-BN/FGT spin valve were respectively measured to be 119% and 20 mT, as shown in Supplementary Fig. 18. Compared to FGT/h-BN/FGT spin valve in our work, the field window under interlayer spacing-modulation were around 16 times larger than that of the vertical FGT/h-BN/FGT spin valves at 5 K, respectively. The experimental results confirmed the great advantage of interlayer spacing modulation in realizing high-performance spin valve devices.
Non-local EB effect in magnetic vdW-structure
Generally, the hysteresis loops of FGT flake remain symmetric without any magnetic EB, unless capped with FPSe18, 19. Nevertheless, the HEB of FGT is also observed at point 2 (Fig. 4a, sample B), a phenomenon that has not been reported yet. Unexpected magnetic hysteresis behavior is observed in connected FGT (position 2), even without LS (Fig. 4b). A 20 mT HEB and a 20 K Tb are observed as shown in Fig. 4b, consistent with FPSe/FGT (position 1). Importantly, the measured HEB, Tb, and HC of connected FGT (position 2) exhibit basically identical reinforcement after LS, as shown in Fig. 4c(Ⅰ) and 4c(Ⅱ). However, we have also measured the hysteresis loops of unconnected FGT (position 3) before and after LS to exclude pressure-induced magnetic hysteresis behavior of FGT, and the results show a lack of the EB effect (position 3) either with or without LS (Fig. 4b(Ⅲ) and 4c(Ⅲ)). In addition, to prove that the phenomenon is not an accidental result, we also fabricated other FPSe/FGT heterostructures containing both connected and unconnected FGT (sample C, see Supplementary Information). As shown in Supplementary Fig. 19, the hysteresis loops of unconnected FGT remain symmetric without any magnetic EB, while connected FGT and FPSe/FGT heterostructure exhibit similar HEB, Tb, and Hc. Figure 4d shows the comparison of HEB and Tb at three positions. Before LS, positions 1 and 2 show the same EB effect, but the EB effect is absent at position 3. After LS, HEB, Tb, and HC show consistent enhancement at positions 1 and 2, and there is still no EB effect at position 3. The temperature-dependent scatter plots of the left (HC−L) and right (HC−R) coercive fields at all three positions are shown in Fig. 4e. The difference between HEB gradually decreases from 73 mT and 68 mT (at 5 K) to 0 mT (at 90 K) for positions 1 and 2, respectively, which are summarized in Fig. 4f. This result indicates that the modulation of HEB and Tb at positions 1 and 2 are basically the same. This non-local coupling mechanism of the horizontal pinning originates from the inherent influence of the FGT. As long as the FGT is connected with FPSe, the scale of transverse propagation can reach tens of micrometers or even hundreds of micrometers, leading to a large horizontal pinning. Due to the limitation of mechanically peeled materials, the exact boundary of the transverse propagation effect has not been studied further. This wide range of EB effect and propagation effects at the micrometer level will be beneficial for the application of industrial devices.
To investigate the non-local coupling mechanism of the horizontal pinning, we propose a model based on the experimental results above, which suggests the possible spin structures in the inset of Fig. 4g and 4h. First, we observed this non-local effect below the Néel temperature of 110 K. Fei et al. pointed out that FGT exhibits single-domain properties at temperatures below 150 K when its thickness falls between 10 nm-100 nm.37 Therefore, within the range of temperature and sample thickness measured in this experiment, FGT can be considered as in the single-domain configuration. When FGT was covered by FPSe, the spin direction of FGT of the heterostructure was pinned by the antiferromagnetic material. Due to the single-domain nature of FGT, the spin of connected FGT was pinned in the same manner, leading to the non-local EB effect. As shown in Fig. 4g and 4h, when no pressure was applied, the pinning effect is not obvious due to the weak interface coupling, corresponding to a small HEB and a low Tb experimentally. When sufficient pressure was applied, the enhanced interlayer coupling led to an increase of the pinning effect, which caused larger coercive fields in the hysteresis loop, as well as larger HEB and higher Tb in the experiments. Moreover, this distance at which EB can be detected in connected FGT can go beyond 40 µm in our samples after LS (Supplementary Fig. 19, where the maximum distance is limited by the length of the sample), showing a simultaneous enhancement between FPSe/FGT heterostructures and connected FGT. This phenomenon can provide a potential platform for studying the experimental determination of spin transport properties. Moreover, understanding of the non-local coupling mechanism of the horizontal pinning will ultimately contribute to the development of the spin logic device adopting 2D FM as a spin transport channel.