Introduction

Antiferromagnetic (AF) spintronics has emerged as a major field in spintronics, fueled by its remarkable attributes of high magnetic field (H-field) stability and ultrafast operation speed1,2,3,4,5,6,7. However, this robustness also leads to a weak response to an external magnetic fields, which hampers conventional magnetic control methods8,9. Synthetic AF materials (SAF), however, offer great opportunities for AF devices due to their relatively weak exchange coupling. SAFs typically consist of two ferromagnetic (FM) layers separated by a metallic or an insulating spacer, where two FM layers are magnetically coupled via Ruderman-Kittel-Kasuya-Yosida (RKKY)-type exchange interaction mediated by nonlocal spin-polarized carriers of the space layer10. In an SAF structure, ferromagnetic or antiferromagnetic coupling can be controlled via the thickness and material of the nonmagnetic space layer11.

Control of the exchange coupling in SAFs shows promise for facilitating fast and efficient magnetization switching in devices. Recent experiments have demonstrated that an electric field can tailor the RKKY coupling and drive transitions between AF and FM12,13,14. These effective modulations typically rely on changes of the Fermi level influenced by charge accumulations at the surface or ion doping by applying a voltage15,16,17,18. Because of efficient charge screening in metallic SAF structures, charge accumulation by the E-field is usually limited to the outmost surface/interface and therefore indirectly affects the RKKY coupling arising from the FM/NM interface. Ion migration into the SAF driven by E-field may somehow cause additional chemical or structural complexity19. Optical control of the RKKY coupling is therefore preferable for its advantages of simpler fabrication, inner-interface access20,21 and flexibility22.

Optical manipulation of the RKKY interaction was initially investigated in semiconductor quantum dot systems and II-VI diluted magnetic semiconductors23,24,25, where the exchange interaction can be established by virtual electron-hole pairs or excitons generated by a sub-band-gap optical excitation26. However, equivalent modulation of the RKKY interaction in an SAF has not been performed yet. Manipulating the interlayer coupling by femtosecond laser pulses is key for ultrafast spin dynamics such as all-optical switching (AOS) where the RKKY coupling is reported to mediate magnetization switching as little as a few picoseconds27,28.

Meanwhile, the interfacial magnetic anisotropy constant is an important parameter in a perpendicular SAF structure. We note that although E-field control of magnetic anisotropy energy (MAE) has been studied in ferromagnets29,30, the systematic consideration of MAE variations during the E-field control of the RKKY coupling, for example in synthetic antiferromagnets, has not been explored, yet it is essential in both fundamental and technical aspects. As they both show interfacial origins, we expect there would be a connection between the manipulation of the MAE and the RKKY coupling.

Here, we use antiferromagnetically coupled [Pt/Co]4/Ru/[Co/Pt]2 as a prototype system to investigate optical manipulations of the RKKY interaction and the interfacial PMA. We show by time-resolved magneto-optical Kerr effect (TR-MOKE) that a fs laser pulse can simultaneously and equally tune the RKKY coupling and the PMA in the [Pt/Co]4/Ru/[Co/Pt]2 system. Upon pump excitations the RKKY strength and the coercivity of a ferromagnetic layer are scaled to the same extent. The modulated RKKY and coercivity shows no dependence on pump-probe delays, excluding contributions of the magnetization amplitude as it varies as a function of the pump-probe delays. This suggests that the optical modulations originate from the Ru space layer which dominates the RKKY exchange coupling. We propose that the pump laser excites conduction electrons of the nonmagnetic Ru layer into a trapped states (e.g., defects or impurities) at the interface leading to a reduction of Fermi wave vector kF of the Ru conduction layer, and therefore minimizing the RKKY coupling between two FM layers. The shift in the Fermi level (EF) associated with this change in the kF alters the electron occupation at the Pt/Co interface and thus the perpendicular coercivity of the constituent FM layer. Our results provide insights into how a femtosecond laser pulse modulates the RKKY interaction and PMA, thereby enabling control of magnetization switching processes in a SAF system for applications in ultrafast magnetic data storage.

Results

Figure 1a illustrates the layer structure of the composite film used in our experiment. The SAF stack consists of Ta(4)/Pt(4)/[Pt(0.6)/Co(0.6)]4/Ru (0.8)/[Co(0.6)/Pt(0.6)]2/Ta(2) (thicknesses in nanometers). The 2 nm Ta capping layer is used to prevent sample oxidation. Here, the unit of [Pt(0.6)/Co(0.6)]4 is read as the bottom ferromagnetic layer, and the [Co(0.6)/Pt(0.6)]2 is set as the top ferromagnetic layer. Figure 1b shows an out-of-plane hysteresis loop measured by a vibrating sample magnetometer (VSM). The top and bottom ferromagnetic layers are antiferromagnetically coupled. The SAF system can be manipulated into four distinct magnetization states of P+, AP+, AP− and P− by an external magnetic field. The blue and red arrows denote the magnetization directions of the top and bottom multilayers, respectively. The bottom [Pt/Co]4 layer is magnetically hard, while the top [Co/Pt]2 layer is magnetically soft (Supplementary Note 1). The antiferromagnetic coupling field Hex can be obtained from the minor loop shift of the top layer. In Fig. 1b, Hup and Hdown in the minor hysteresis loop indicates the switching field in the up and down sweeping directions of the external field, respectively. The minor loop displays a substantial net positive shift of 3.29 kOe (referred to as the antiferromagnetic coupling field Hex) obtained via (Hup + Hdown)/2. Hex is proportional to the RKKY exchange coupling. Meanwhile, the coercivity of the [Co/Pt]2 soft layer can be represented as HC = (Hup − Hdown)/2, showing a magnitude of 0.24 kOe.

Fig. 1: SAF-sample structure and the static magnetic properties.
figure 1

a SAF sample schematics. b Static out-of-plane hysteresis loop measured by VSM. The green and orange lines represent the process that sweeps the magnetic field from negative to positive and from positive to negative, respectively. The switching fields of the minor loop in two sweeping branches are defined as Hup and Hdown, respectively. The RKKY coupling field Hex and the coercivity HC are extracted from Hup and Hdown.

Full size image

Figure 2a displays transient minor hysteresis loops measured at the time delay where the largest demagnetization amplitude is reached for different pump fluences. The minor loop is measured in a field range where the magnetically hard bottom layer is magnetically fixed after an initial positive saturation of the whole system. At a low pump fluence of F = 0.96 mJ/cm2, the switching fields Hup and Hdown are comparable to their counterparts in the case of without pump pulses. At higher pump fluences of F = 3.20 mJ/cm2 and 5.60 mJ/cm2, Hup shows a noticeable decrease while Hdown remains unchanged. To quantitatively analyze the effect of pump fluence on minor loops, we plotted in Fig. 2bHex and HC as a function of pump fluence. Hex decreases from 3.16 kOe to 3.02 kOe, while HC displays a decline from 0.23 kOe to 0.07 kOe. They approximately show the same reduction of magnitude, namely ΔHex ≈ ΔHC = 0.16 kOe. This suggests an intrinsic coherence of the optical modulations on Hex and HC by increasing the pump fluence. Similar results in a series of samples with different RKKY strength by changing the Ru thickness and the [Co/Pt] repetition number are observed (Supplementary Note 2), confirming the reliability and the consistency of our results.

Fig. 2: TR-MOKE measurements.
figure 2

a Transient minor hysteresis loops at selected pump fluences. As guided by the dashed lines, Hup shows an obvious reduction at high pump fluences while Hdown remains unchanged. b Calculated Hex and HC values from the minor loops as a function of pump fluence F. Within experimental error bars they show the same trend after fs excitations. c Minor hysteresis loops probed at three different pump-probe delays under a fixed pump fluence of 5.60 mJ/cm2. d Extracted time-delayed Hex and HC from transient loops in (c), showing negligible variations within 10 ps. Error bars indicate the uncertainty in determining the displayed value from the experimentally measured data.

Full size image

We notice that the Hex and the HC show similar and weak dependence on the pump-probe delays. Figure 2c displays “snapshots” of the transient minor hysteresis loops the time delays of Δt = 0.063 ps, 0.630 ps, and 9.763 ps (Supplementary Note 3 determines the time zero by fitting the demagnetization curve). The pump fluence is fixed at F = 5.60 mJ/cm2. The amplitude of the transient minor hysteresis loops corresponds to the pump-induced demagnetization amplitude. It is notably diminished at Δt = 9.763 ps, indicating a substantial recovery of the magnetic order. Within the observed time window of the demagnetization process (~10 ps), the extracted Hex and HC show no pump-probe dependence except at Δt = 0.63 ps as shown in Fig. 2d whose y-axis ranges (~0.2 kOe) are the same as those of Fig. 2b. The amplitudes of the small spikes at Δt = 0.63 ps in Fig. 2d are much smaller than ΔHex and ΔHC. The transient magnetization as shown in Fig. S4b drops first and then recovers within a few ps. Meanwhile, the extracted Hex and HC at different time delays as included in Fig. 2d remain stable. This suggests that the changes of Hex and HC by a fs light pulse are not correlated with transient magnetization amplitude.

It has been demonstrated that a fs laser pulse can excite the domain nucleation and the domain wall propagation due to circular dichroism31, which may contribute to the variations of magnetization switching behavior. By changing circular polarization of pump beam, we have found that both the HC and Hex in our system show no dependence on circular polarization of the pump beam as detailed in Fig. S10a, b. This indicates that the possible light-induced domain wall propagation does not take effects here.

We have further considered whether thermal effects could lead to the concomitant reductions of Hex and HC. The thermal heating is an unavoidable factor when the pump fluence is increased. We recorded the minor loops as a function of the temperature by both the SQUID in Fig. 3a, b and the MOKE measurements in Fig. 3c, d. In Fig. 3a, as the temperature T increases from 300 K to 380 K, Hup and Hdown decreases simultaneously and the minor loop shifts towards the negative-field direction. It is revealed that Hex shows a reduction of ~ 0.4 kOe from 3.22 kOe at 300 K to 2.82 kOe at 380 K. In parallel, HC shows only a marginal fluctuation of ~ 0.05 kOe. We further confirm this behavior (ΔHex > ΔHC) through MOKE measurements by changing the static temperature. The employed pump fluence is fixed at F = 2.40 mJ/cm2. As shown in Fig. 3d, Hex decreases by 0.4 kOe while HC stays constant at ~0.08 kOe for a temperature increase of 80 K. The SQUID results can exclude possible contributions of the inhomogeneous weights of the individual layer due to the optical penetration depth or multilayer reflections in the MOKE.

Fig. 3: Temperature-dependent minor hysteresis loops.
figure 3

a Temperature-dependent minor hysteresis loops by SQUID. Both Hup and Hdown are changed with the static temperature. b Extracted Hex and HC from (a) as a function of temperature. Hex decreases with the increase of the static temperature while HC stays unchanged within experimental error bars. c Temperature-dependent minor hysteresis loops by MOKE at a fixed pump fluence F of 2.40 mJ/cm2. As in (a), similar shifts of Hup and Hdown are shown. d Extracted Hex and HC from (c). The complementary magneto-optical method again confirms that a temperature increase of 80 K only affects Hex, but not HC. Error bars indicate the uncertainty in determining the displayed value from the experimentally measured data.

Full size image

As shown in Fig. 3b, d by increasing the static temperature, Hex varies even for a temperature increase of 20 K. This suggests that Hex shows very sensitive dependence on thermal effect. According to the three-temperature model, transient lattice temperature typically increases to an equilibrium at a few ps after fs excitations32,33. However, the extracted time-delayed Hex and HC in Fig. 2d do not change up to 10 ps, indicating that within this timescale there is no noticeable transient temperature rise. In addition, the TR-reflectivity ΔR/R (Fig. S9 in Supplementary Note 9) shows a negative dip within 1 ps, indicating that nonthermal electrons are scattered before they couple to the lattice system. If thermal heating plays a role, the ΔR/R signal should show a growing positive component, which represents the lattice temperature, by increasing pump fluence as reported by Shim et al.34. However, the normalized ΔR/R in Fig. S9 are all negative with no fluence dependence. It has been shown in a multilayer structure that transient electron temperature decays rapidly to the ambient temperature within 1 ps before it couples to its lattice temperature35. Therefore, it can be concluded that any possible thermal effects do not play a major role for the concomitant variations of Hex and HC upon fs excitations.

Discussion

As discussed above, the nonchanged Hex at various pump-probe delays suggests that upon fs excitations the magnetization magnitude has a minor impact on the RKKY interaction. The RKKY interaction can be described by the following effective Hamiltonian36

$$H=-{\sum}_{i,j} \, J\left({R}_{ij}\right){S}_{i}\cdot {S}_{j}$$
(1)

where \({R}_{ij}=\) \({R}_{i}-{R}_{j}\); \({S}_{i}\) is the spin operator on the atom \(i\) and \(J\left({R}_{ij}\right)\) is the exchange interaction between the neighboring spins on sites \(i\) and \(j\). This guides us to relate the RKKY reductions to the modulation of \(J\left({R}_{ij}\right)\) by the fs laser pulses.

In a metallic multilayer sample, when two magnetic thin films (referred to as S1 and S2) are separated by a non-magnetic spacer layer with a thickness of d, the interlayer exchange coupling constant per unit area at the interface can be defined as J12. This constant oscillates in both the sign and the magnitude within the spacer layer and decays as 1/z2, where z represents the distance from a selected reference layer within the thin films. In the limit of a large distance z, J12 takes on a simple form36:

$${J}_{12}\left(z\right)={J}_{0}\frac{{d}^{2}}{{z}^{2}} sin \left(2{k}_{F}z\right)$$
(2)

where J0 is a constant, and kF represents the Fermi wave vector. In our experimental setup, the thickness d of the spacer layer remains constant. Therefore, the principal focus must be on the variations of the Fermi wave vector kF by fs laser pulses, which can affect the RKKY interaction as well as shown by Eq. 2.

We accordingly propose a diagram in Fig. 4a to explain the simultaneous reductions of the RKKY interaction and the PMA by fs laser pulse excitations. When the sample is irradiated with a fs laser pulse, it can introduce nonequilibrium electron distributions at the Co/Ru interface. When trapping states (such as defects or impurities) exist above the Fermi level EF, photoexcited holes near the EF will cause a downward shift of the EF, thus leading to a new EF’. These trapping states allow much longer lifetimes of the photoholes than the 10 ps timescale that we have observed. As shown in Fig. S9, the initial negative ΔR/R dip indicates absorptions of photoexcited electrons/carriers, showing the existence of trapping states. The drop in photocarriers consequently introduces a deformation potential, a manifestation of carrier-density change. This launches a picosecond strain wave pulse37,38, which can drive coherent phonon oscillations.

Fig. 4: Schematic diagram of optical control of the PMA and the RKKY.
figure 4

a Diagram of how fs-laser excitations modulate the RKKY coupling and the PMA in a SAF system. Upon fs excitations, the pump photon energy (1.55 eV) excites electrons in Ru (indicated by the red vertical arrow) from occupied states below the Fermi level EF to unoccupied trapping states above EF, resulting a down-shift of the Fermi level from EF to EF’. Considering the parabolic band dispersion at the Ru interface, a downshift of EF can result in a decrease in the RKKY coupling. Meanwhile, at the Co/Pt interface such a ΔEF shift will cause electron redistributions between dxz and dyz orbitals due to a smaller energy gap, thus causing a concomitant reduction of the PMA. b Transient hysteresis loops recorded at 0.063 ps under different pump fluences for a thin [Co/Pt]4 film. c Extracted HC values from the [Co/Pt]4 thin film and from the soft layer in the SAF sample as a function of pump fluence. Here, the [Co/Pt]4 stack is inserted in a sample structure of Ta(4)/Pt(4)/[Pt(0.6)/Co(0.6)]4/Ta(2) without the RKKY coupling. Error bars indicate the uncertainty in determining the displayed value from the experimentally measured data.

Full size image

The oscillation period lasts for ~75 ps, suggesting that trapping states remain stable within such a few tens of ps. This is consistent with the fact that HC and Hex do not change as a function of time delays within 10 ps. By increasing pump fluence, HC and Hex change clearly, in accordance with the fact that the trapping states can be more filled upon larger pump fluence.

The Ru space layer is a nonmagnetic metal, whose conduction electrons can be described by the free-electron dispersion model of \(E={\left(\frac{\hslash k}{2m}\right)}^{2}\) Consequently, as depicted in the right-lower part Fig. 4a, the decrease of EF proportionally leads to a reduction in the Fermi wave vector kF, and thus to a decrease of the RKKY interaction (or equivalently the Hex). The photocarrier lifetime of hundreds of picoseconds have also been observed in other magnetic heterostructures39.

Using the lattice structure in Fig. 5a, the light manipulation of the AF coupling can be understood by the first-principle calculations using a Green function method with TB-LMTO40 (Supplementary Note 4). Electron transitions in metals by a laser can result in a corresponding decrease in the Fermi energy22,41. We can qualitatively reveal the light-manipulated AF coupling as a function of the Fermi energy. The calculated AF coupling ( Jex) versus Fermi energy shift (E − EF) is shown in Fig. 5b. It can be observed that as the Fermi energy decreases, the AF coupling continues to decrease. The theoretical calculations are consistent with the above proposed model.

Fig. 5: Influence of Fermi level shift (E − EF) on AF coupling Jex by First-principle calculations.
figure 5

a The sketched lattice interface of the SAF structure for calculations. b The calculated AF coupling Jex as a function of the shift of Fermi level.

Full size image

In 3d ferromagnetic metal/noble metal, the PMA is demonstrated to originate from the interface hybridizations and is sensitive to electron filling42,43,44,45. At the same time, the coercivity has been found being linked to the changes of the magnetic anisotropy46,47 in the case of either the single-domain or the nucleation and domain wall propagation switching model. It is reported that the hybridization between Pt 5d and Co 3d at the interface in Co/Pt multilayers enhances the orbital magnetic moment, thus the PMA due to spin orbital interaction (SOC)42. In specific, the SOC matrix elements with the same spin \(\langle {d}_{{xy}}|{H}_{{SO}}|{d}_{{x}^{2}-{y}^{2}}\rangle\) and \(\langle {d}_{{xz}}|{H}_{{SO}}|{d}_{{yz}}\rangle\) contributes to PMA42,44. \({d}_{{xz}}\) and \({d}_{{yz}}\) orbitals are split by SOC with an energy gap of 70 meV. While \({d}_{{xy}}\) and \({d}_{{x}^{2}-{y}^{2}}\) are located by the crystal field with an energy separation much larger than 70 meV45. Thus, electron occupations between \({d}_{{xz}}\) and \({d}_{{yz}}\) orbitals at the Co/Pt interface are more easily to be modulated by perturbations such as optical excitations.

As depicted in the right-upper part of Fig. 4a, ΔEF occurs simultaneously at the Pt/Co interface (also at the Co/Ru interface as mentioned above), leading to electron redistributions therein, for example, between \({d}_{{xz}}\) and \({d}_{{yz}}\) orbitals. This can diminish the PMA, similar to previous reports of E-field control on PMA29,30,48. In the absence of the interlayer interaction for a [Co/Pt]4 thin film (Fig. 4b), its HC (red squares in Fig. 4c) shows a step-like decrease as a function of F, while the HC in the SAF (blue squares in Fig. 4c) declines continuously. This contrast suggests that the deviations of HC in a SAF structure is caused by Hex. Therefore, when Hex is modulated by fs laser excitations due to ΔEF, HC is changed consecutively.

In our experiments, the RKKY field Hex shows a drop by 4.9%. This corresponds to an EF shift of ~35 meV according to our theory calculations in Fig. 5b. Such a shift is large enough to cause the electron redistributions among \({d}_{{xz}}\) and \({d}_{{yz}}\) orbitals with a gap of 70 meV, and thus modifying PMA.

The above proposed explanation involves a photoexcitation process, which means that changing the pump excitation may give rise to different modulations on both Hex and HC. This is verified by exchanging the pump and probe wavelengths, namely, exciting the sample with a photon energy of 3.1 eV while probing the dynamics with a photon energy of 1.55 eV. As shown in the Fig. S5 (Supplementary Note 5), Hup of the transient minor loops excited by the 400 nm laser show an obvious decrease at F = 0.48 mJ/cm2, and remains almost stable after F = 0.80 mJ/cm2. For the 800 nm excitations in Fig. 2, variations of Hup actually occur after F = 1 mJ/cm2. Such an efficient excitation by a 400 nm pulse is supported by the measured optical reflectivity spectrum. As shown in Fig. S6c, SAF reflectivity at 800 nm is larger than that at 400 nm, showing larger absorption at 400 nm. More efficient excitations by 400 nm are likely due to the generated non-thermal electrons at higher energy levels. If the absorbed-photon amounts play a role, the transient MOKE loops at higher pump fluence pumped by 800 nm would be similar to the MOKE loops pumped by 400 nm laser at a low fluence (see also Supplementary Note 5). However, the shapes of transient loops excited by 800 nm and 400 nm are clearly different. Thus, nonthermal electrons contributions to the efficient excitations are preferred.

Here, the optically modulated RKKY and PMA assisted by trapping states have been found to affect the switching process of SAF magnetization (Fig. S7) and magnetization precession dynamics (Fig. S8), which may have potential ultrafast spintronic applications. Recent research on a PN-Si/CoFe/Cu/CoFe sample demonstrates that light illumination induces a switch between FM and AFM states49, where there is an exchange of photoexcited electrons or carriers between the SAF layer and the substrate. In our current study, the naturally oxidized SiO2 layer on a Si substrate prevented the generation and propagation of any photoexcited electrons or carriers. With a combination of a SAF structure and a clean semiconductor surface, the RKKY and PMA may be optically modulated more profoundly. However, such a study is beyond the scope of this work.

In summary, our investigation of the transient hysteresis loops in a synthetic antiferromagnetic structure shows optical modulations on the RKKY interaction and the MAE by a femtosecond laser pulse. By increasing the pump fluence, simultaneous reductions of both the Hex and the HC are observed. The results are explained in terms of changes in the electronic DOS and orbital occupancy at the Co/Ru and Pt/Co interfaces. The fs-pulse excitation induces a downshift of the Fermi level EF within the Ru layer, subsequently reducing the Fermi wave vector kF and then resulting in a continuous decline of the RKKY coupling. Meanwhile, the Fermi level shift lunches subtle electron transfer at the Pt/Co interface between \({d}_{{xz}}\) and \({d}_{{yz}}\) orbitals, leading to a decrease in the PMA. As expected from our proposed model, different pump excitation process (400 nm vs 800 nm) is shown to have an impact on the modulations. Our results achieve, for the first time, optical control of the RKKY coupling and the PMA by a femtosecond laser pulse and provide fundamental understandings on this. The findings are expected to play an important role in the energy-efficient magnetization switching.

Methods

Sample preparation and VSM measurement

The sample was grown via high-vacuum magnetron sputtering on Si substrates with a native oxide. The base pressure is maintained at less than 1 × 10−7 Pa. The static magnetic properties were measured by a vibrating sample magnetometer (VSM) at room temperature.

Time-resolved MOKE/reflectivity measurements

The magnetization dynamics were investigated using the Time-Resolved Magneto-Optical Kerr Effect (TR-MOKE) technique, employing a pulsed Ti:sapphire regenerative amplifier with a central wavelength of 800 nm, a pulse duration of approximately 50 fs, and a repetition rate of 1 kHz. Dynamic behaviors were induced by an intense pump pulse laser at 800 nm, while the transient MOKE signals were detected by a time-delayed probe beam at 400 nm with a much lower intensity. Another beam splitter was placed in the probe beam path, before it was reflected from the sample, to obtain the reference beam and probe pulses for TR-reflectivity measurements. The pump beam was incident on the sample with a 400 µm diameter spot size, while the probe beam was angled approximately 4° away from the normal direction of the film plane, with a spot size of 250 µm in diameter. The external magnetic field was applied perpendicular to the sample surface. In this configuration, the Kerr rotation is sensitive only to changes in the out-of-plane magnetization component. To precisely determine the time zero, please refer to Supplementary Note 3 in the Supporting Information.