The present invention relates to memory storage systems, and particularly to a memory storage system that uses current to move magnetic domain walls in a magnetic racetrack.

Background

Racetrack memory devices are gaining interest as high-density storage devices. These devices are disclosed, for example, in US-B 6,834,005, More advanced racetrack memory cells have already been developed including vertical nanowire storage elements, as disclosed e.g. in US-A 2014/204648.

Especially, racetrack devices based on chiral domain wall (DW) magnetic bits in perpendicularly magnetized ferromagnet/heavy metal thin film systems are a promising candidate for next generation spintronic memories (US-A 2014/0009994, US-A 2014/0009993). These currents can be used, in particular, to manipulate magnetic bits that are encoded within magnetic spin textures (domains, skyrmions, or antiskyrmions) in nanoscale racetracks.

Current-induced domain wall motion (CIDWM) has significantly evolved from in-p!ane magnetic to synthetic antiferromagnetic (SAF) racetracks due to advances in volume spin-transfer torque (STT) and spin-orbit-torque (SOT) mechanisms. Driven by a chiral spin torque that arises from the spin-orbit coupling in the presence of broken inversion symmetry at ferromagnet/heavy metal (HM) interfaces, Neel domain walls in thin films with strong perpendicular magnetic anisotropy (PMA), stabilized by a Dzyaloshinskii- Moriya interaction (DMI) at the ferromagnet/HM interfaces, can be moved along the current direction at high velocities (e.g. EP3171364A1), in both straight and curved racetracks. The fast and energy efficient motion of such magnetic bits along 2D or 3D racetracks by spin current is a key challenge for its commercial implementation.

An even more efficient DW motion was reported in synthetic antiferromagnet (SAF) racetracks that are composed of two perpendicularly magnetized ferromagnetic subracetracks coupled antiferromagnetically across an ultrathin ruthenium layer. The giant exchange coupling torque (ECT) in the SAF structure provides an additional dominating driving mechanism that allows for an increased DW propagation velocity beyond ~1,000 m/s. The ECT in rare earth-transition metal alloys is further maximized 20/004 MAX at the angular momentum compensation temperature of the ferrimagnetic alloy. Recently, efficient CIDWM was also found in certain magnetic insulators.

Significant progress has been made regarding a detailed understanding of the interface derived chiral spin torque and magnetic chirality with respect to the underlying mechanisms of CIDWM, for example, by varying the HM layer that is in contact with the interface ferromagnetic layer or by tuning the thickness of the ferromagnetic layers.

Object of the Invention

However, the domain wall velocity is still too low and the threshold current densities are still too high for commercially feasible fast and low power racetrack memory devices. Accordingly, it was an object of the present invention to provide for a more efficient current-induced domain wall motion.

Brief description of the invention

The present invention significantly reduces the threshold current densities and greatly increase the efficiency of domain wall motion by introducing an atomically thin 4 d or 5 d preferably 4d metal “dusting" layer (DL) at the ferromagnetic/heavy metal (HM) interface. In a further preferred embodiment, a sub-atomic-iayer-thickness dusting layer of Pd and Rh, more preferably a dusting layer of a thickness of just one monolayer at the H M/ferromagnetic interface is introduced, which increases the domain wall’s velocity by a factor of up to 3.5 at a given current density. The Neel DWs move more than three times faster, for the same injected current density, compared to otherwise identical structures without any DL. Moreover, the threshold current density, Jt _h , defined as the minimum current density required to overcome the effective pinning field and move the DW, is substantially reduced by incorporating atomically thin DLs.

Without wishing to be bound by this theory it is believed that this improvement is due to a subtle interplay of tailored spin-orbitronic parameters, i.e. parameters that originate from the spin-orbit coupling effects; specifically, the Dzyaloshinskii-Moriya interaction and the uniaxial magnetic anisotropy. The present invention shows how significant interfacial modifications are, to allow for tailored racetracks with enhanced efficiency of chiral domain wall motion and it directly demonstrates the close inner correlation of the Dyzaloshinskii-Moriya interaction with the uniaxial anisotropy. Brief description of the Drawings

Figure 1 shows Engineered FM and SAF Racetrack Memory structures with interfacial dusting layers. a Schematic representation of the FM (upper panel) and SAF (lower panel) racetrack structures with a dusting layer (DL) inserted between the heavy metal

(Pt) and ferromagnetic metal (Co) layers. b Upper panel: schematic illustration of the atomic stack of the FM/DL/FIM structure along the fee (111) direction; lower panel: the color-coded elements employed in the FM/DL/HM stack; orange, green, and light blue correspond to FM, DL, and FIM, respectively. The numbers in the upper right position of each element square correspond to (from top to bottom): the spin-orbit coupling constant, spin diffusion length, and Stoner criterion parameter, respectively. The number in the lower left position is the in-plane lattice constant of the corresponding fee (111) unit. c Scanning electron microscopy of a typical racetrack. d Cross sectional FIRTEM image. e The corresponding EDX mapping of a FM film with 0.1 nm thick Pd dusting layer, in which the presence of the atomically thin Pd layer is highlighted. Figure 2 shows the Interfacial DL engineered chiral domain wall motion in FM and SAF structures. a-f current-induced DW motion in the FM (left panel) and SAF (right panel) structures with various DL materials: Pd (a and b, orange background), Rh (c and d, violet background), and Ir (e and f, blue background). The insets in (a) and (b) illustrate typical Kerr images of the domain wall motion in response to a series of injected current pulses (~ 1.0 c 10 ⁸ A/cm ² ) composed of, respectively, twelve (a) and four (b) 10 ns pulses in the FM and SAF samples with 0.1 nm Pd DL, which confirms that all the DWs move in the direction of current injection. Differences in image contrast originate from the magnetization difference in the FM and SAF samples. The bright and dark parts correspond to down ) and up (Q or t ) domains. The thickness of the inserted dusting layers are varied from 0 nm (navy squares), 0.1 nm (red circles), 0.2 nm (orange diamonds), 0.3 nm (light blue triangles), 0.4 nm (olive triangles), 0.5 nm (purple triangles) and 0.7 nm (orchid triangles). Figure 3 shows DL thickness dependence of the DW velocity (v) and threshold current density (J _th ). a and b show the DW velocity at a fixed current density (~ 1.2 x 10 ⁸ A/cm ² ) in the FM and SAF samples with various DL thicknesses (toi). The dash lines represent the velocity of the reference samples and the stationary state of the domain walls. The right axis is the normalized DW velocity (vnor) with respect to the reference samples. c and d are the threshold current density (J _th ) as a function of different DL thicknesses in the FM and SAF samples. The dashed lines represent the threshold current density of the reference samples. The right axis represents the normalized threshold current density (J _th _nor) with respect to the reference samples. The colored regions illustrate the range of interfacial DL thickness where the efficiency of CIDWM is maximized. Pd, Rh, and Ir DL correspond to orange squares, violet circles and light blue diamonds, respectively.

Figure 4 shows the Longitudinal magnetic field dependence of DW velocity on dusting layer. a-f the longitudinal field dependence of DW velocity (v-H _x ) at a fixed current density (~ 1.2x 10 ⁸ A/cm ² ) in the FM (left panel) and SAF (right panel) structures. The open and filled symbols represent the domain configurations, respectively. The samples with DL thicknesses of 0.1 nm (a and b), 0.2 nm (c and d), 0.3 nm (e and f) and 0.5 nm (g and h) are shown. The reference samples and those with Pd, Rh, and Ir DL are represented by navy triangles, orange squares, violet circles, and light blue diamonds, respectively.

Figure 5 shows the Interfacial dusting layer engineered magnetic properties. a The DL thickness dependence the effective uniaxial anisotropy constant in the FM structures. b vig in SAF structures. c DMI constant D calculated from the FM cases. d the DMI constant D of all samples plotted as a function of in which the dashed lines are the linear fitting of the 4 d elements DL cases (Pd and Rh, green) and 5of elements DL cases (Ir and Pt, red), respectively. The reference samples and those with Pd, Rh, and Ir DL are represented by navy triangles, orange squares, violet circles, and light blue diamonds, respectively.

Detailed description of the invention

The basic structure of a racetrack device according to the present invention is based on a ferromagnetic (FM) structure or a synthetic antiferromagnetic (SAF) structure.

The ferromagnetic structure comprises one or more, preferably two or three layer(s) of a ferromagnetic material. If the ferromagnetic structure comprises more than one layer, preferably two neighboring layers are not identical. The ferromagnetic layer is made of ferromagnetic material selected from one or more of:

Fe, Co, Ni or Mn, or an alloy of Fe and/or Co, or an alloy of Ni that may further include one or more of Fe and Co, or an alloy of Mn that may optionally further include one or more of Fe, Co or Ni, or a compound including Fe and at least one of ir, Al, Si or Ge, and a compound including Mn and Si.

Typically, the ferromagnetic structure may have a total thickness in the range of 0.5 to 1.5 nm, preferably 0.75 nm to 1 .3 nm, more preferred 0.9 nm to 1.2 nm. Each individual layer of the ferromagnetic structure may - independently of one another - have a thickness in the range of 0.1 nm- 1.5 nm, preferably 0.12 nm to 1.0 nm, more preferred 0.14 nm to 0.8 nm.

The ferromagnetic layer(s) of the ferromagnetic structure are preferably sandwiched between an FIM layer and a coupling layer.

The FIM layer includes at least one of Pt, Ir, W, Ta or Ru. The FIM layer may advantageously have a thickness between 0.8 nm- 2.0 nm, preferably 1.0 nm to 1.8 nm, more preferred 1.2 nm to 1.7 nm.

The coupling layer includes at least one of Ru, W, Ta or Ir. The coupling layer may advantageously have a thickness between 0.4 nm 1.5 nm, preferably 0.6 nm to 1.0 nm, more preferred 0.7 nm to 0.9 nm.

The synthetic antiferromagnetic structure may be comprised of two ferromagnetic layers coupled antiferromagnetically via a coupling layer. In a preferred embodiment, the FM/coupling layer/FM sandwich is deposited on a FIM layer. The ferromagnetic layers, the coupling layer and the FIM layer are as described above. Preferably the coupling layer is comprised of Ru or Ir.

According to the invention the FM and/or SAF Racetrack Memory structures comprise an interfacial dusting layer. This dusting layer is made of a 4 d or 5 d metal, preferably a 4d metal and most preferred a metal with a long spin-diffusion length of preferably more than 5 nm. It is further preferred that the metal exhibits an fee structure. Most preferred are Pd and Rh, while Ir and Ru are less preferred.

The dusting layer is preferably located at the HM/FM interface, i.e. the dusting layer is preferably sandwiched between the HM and FM layer.

The thickness of the dusting layer is preferably in the range of 0.1 to 1.5 nm, more preferred 0.1 to 1.0 nm and most preferred 0.2 to 0.7 nm. This means that the dusting layer may have a dimension which is in the sub-atomic-layer-thickness range, which is defined as being within or even below the lattice constant of the selected DL material. In some embodiments the dusting layer has a thickness which is in the range of about the diameter of an atom of the DL material, i.e. in the range of about 0.2 nm; this thickness is called “one monolayer”. The “sub-atomic-layer” thickness as well as the “one monolayer” thickness is the equivalent thickness that is estimated from the sputtering deposition rate.

The individual layers of the racetrack device, including the dusting layer, can e.g. be deposited by magnetron sputtering preferably at room temperature on silicon wafers, which are preferably thermally oxidized so that they are covered with a SiC>2 layer (~ 30 nm). The layers are preferably sandwiched between a bottom TaN layer (~ 2 nm) and a capping TaN layer (~ 5 nm) both with high resistivities. The deposition parameters of these materials can e.g. be calibrated by quartz crystal microbalance and X-ray reflection. The layer thickness can be determined and controlled by the amount (gram-atom) of material sputtered per square unit (e.g. nm ² ). Racetrack nanowires can be fabricated using photolithography and argon ion milling. The domain walls (DWs) can be created in the racetrack nanowires by injecting pulses of current in the presence of external longitudinal magnetic fields.

Application

The present invention can be applied in many different areas of technology, e.g. spintronics, including but not limited to: magnetic random access memories; magnetic recording hard disk drives; magnetic logic devices; security cards using magnetically stored information; semiconductor devices wherein large magnetic fields provided by domain wall fringing fields can be used to locally vary the electronic properties of the semiconductor or semiconductor heterostructure; mesoscopic devices, which are sufficiently small so that the electronic energy levels, therein, can be substantially affected by the application of local magnetic fields; etc..

Advantages of the invention

In models of an ideal racetrack with a homogenous film stack there is no threshold current for CIDWM. Yet, in reality a threshold current of J _th has to be applied in order to depin the DW, by thermally aided excitations across an energy barrier that hampers the DW from moving freely in the wire. Once certain DLs are introduced this barrier is decreased, as indicated by reductions in the uniaxial anisotropy energy Ku and coercive field Hc: thus a decrease in J _th is observed. However, when the DL is further thickened, an inevitable decrease in θ ^SH gives rise to a subsequent increase in J _th . It is important to notice that for the SAF case, J _th corresponds well to that of the corresponding FM case, which suggests the dominant role of the LM layer in determining J _th .

In order to make possible applications of racetrack memory devices based on CIDWM, both low J _th and high DW velocity are often needed. For the FM case, the linear variation of on D (=Dzyalonshinskii-Moriya interaction co-efficient) means that once a smaller J _th is realized, the saturation velocity ( v _D = yD/M _s ) is also smaller. However, in the SAF case, since the DW is mainly driven by the ECT, the maximum velocity is largely determined by the exchange coupling constant when the two sub-layers are the same with each other. Thus, a low J _th and high velocity can both be achieved in the ECT-driven DWM of SAF structure by simply decreasing as indicated from the DL thickness dependence of vi _g , as shown in Fig. 5d. When Pd and Rh DL are inserted into the system, the total 0 ^SH is well preserved due to their relatively long spin diffusion lengths, see Fig. 1b. Hence, the considerable enhancement of v _ig comes mainly from the increase of DW width due to the decrease of For the Ir case, however, just as in the FM case, the quickly dropping θ ^sH can no longer provide enough SOT for the DWM. Thus, a v _ig with a smoothly decreasing trend is observed.

In conclusion, the present invention provides a novel method to substantially decrease the current needed to both depin and to move chiral domain walls in magnetic racetracks. This method involves the insertion of atomically thin 4 d or ScFelement dusting layers, preferably dusting layers made from 4d-elements with fee structure at critical interfaces, preferably at the FM/HM interface in magnetic multilayers that form simple ferromagnetic or synthetic antiferromagnetic racetracks. A clear linear correlation between the perpendicular magnetic anisotropy exhibited by the ferromagnetic racetracks and the Dzyaloshinskii-Moriya interaction that gives rise to the chirality of the domain walls was found for both 4d and 5d elemental insertion layers. These findings are realized through the controlled manipulation of interfaciai spin-orbit coupling

The present invention is illustrated by the following examples.

Examples

Atomically thin dusting layers

Two sets of structures were prepared by DC magnetron sputtering at room temperature as shown in Fig, 1a: a ferromagnetic structure consisting of Co (0,3 nm)/Ni (0.7 nm)/Co (0.15 nm) sandwiched between a Pt (1.5 nm) HM layer and a Ru (0.85 nm) coupling layer, hereafter referred to as a FM structure; and a synthetic antiferromagnetic structure deposited on the same Pt (1.5 nm) HM layer and consisting of a lower ferromagnetic layer of Co (0.3 nm)/Ni (0.7 nm)/Co (0,15 nm) and an upper ferromagnetic layer of Co (0.5 nm)/Ni (0.7 nm)/Co (0.15 nm) antiferromagnetically exchange coupled through a Ru (0.85 nm) coupling layer, hereafter referred to as a SAF structure. A series of atomically thin layers (hereafter referred to as dusting layers, DL) of Pd, Ir, Rh, and Ru with thicknesses varying from 0.1 to 0.7 nm are inserted directly onto the Pt HM layer in both structures before depositing the ferromagnetic materials. Schematic images of the FM and SAF structures are shown in Fig. 1a with the elemental dusting layers illustrated in Fig. 1b. CIDWM was studied in a typical racetrack that was 3 pm wide and 50 pm long, fabricated by photolithography and Ar ion milling (Fig. 1c). The motion of individual DWs in these nanowires in response to voltage pulses of a fixed length (~10 ns) was detected using Kerr microscopy. The DW positions in the nanowire before and after the pulse injection are recorded and, thereby, used to determine the DW velocity along the racetrack.

A cross-sectional high-resolution transmission electron microscopy (HRTEM) image of the FM structure with a 0.1 nm palladium dusting layer is shown in Fig. 1d. The image presents a highly (111) oriented structure of the face-centered cubic (fcc) thin film structure along the out-of-plane direction. The very smooth surface of the films is confirmed by atomic force microscopy imaging. The high-angle annular dark field scanning TEM (FIAADF-STEM) image and the associated energy dispersive X-ray spectrometry (EDX) maps of the layered structure in Fig. 1e directly reveal the Pd dusting layer at the expected location between the Pt layer and the Co/Ni/Co layer, even though the inserted Pd DL is only 0.1 nm thick.

Current induced chiral DW motion

The chira! spin torque drives DWs along the direction of injected current irrespective of the DL parameters. Distinct CIDWM behaviors are observed in the FM and SAF structures that depend sensitively on the DL material and thickness. In the Pd DL case, the threshold current density Jt _h required to observe DW motion is found to be significantly decreased in the FM structure with a Pd DL as thin as only 0.1 nm (Fig. 2a), and the DW velocity is increased over the entire range of current density considered. The maximum current density that can be applied to the racetracks is limited by the formation of multiple magnetic domains that is believed to be due to an increase in temperature of the nanowire as has previously been observed in nanowires of comparable resistance. As the Pd layer thickness is increased, the maximum current density decreases together with a degradation of the CIDWM performance. By contrast, for the SAF structure, Fig. 2b shows that the introduction of Pd DLs improves the efficiency of the CIDWM for all DL thicknesses considered, with significantly higher DW velocities for otherwise the same current density compared to the SAF structure without any DLs (reference SAF). The efficiency of the DW motion is maximized for Pd DL thickness that are -0.2 nm thick with a DW velocity of up to -1000 m/s, as shown in Fig. 2b. This velocity is up to 3.5 times higher by comparison with the same structure without the Pd DL at the same current density. The racetracks with Rh dusting layers behave similarly as for Pd DLs. The range of DL thickness for which the CIDWM is enhanced is extended up to 0.4 nm for the FM case and with a substantially reduced J _th (Fig. 2c). The PMA for a 0.7 nm Rh DL was too weak to allow for CIDWM measurements. For the SAF case shown in Fig. 2d, the CIDWM is enhanced for 0.1 and 0.2 nm thick Rh DLs. For thicker Rh layers (>0.3 nm), however, the DW velocity rather drops with increasing current density. Though the PMA is so weak at 0.7 nm in the FM case that no DW motion could be successfully detected, stable single DWs can still be generated in the SAF case as a result of the exchange coupling, but no sizeable CIDWM could be observed before thermal nucleation of random DWs takes place. A significantly different behavior is observed when an Ir DL is employed. In the FM case, as illustrated in Fig. 2e, Jt _h drops when the Ir layer is 0.1 nm thick, but then increases almost linearly with further increases in the Ir thickness. Only a very slight enhancement in the CIDWM velocity was observed at low current densities for 0.1 nm Ir but otherwise CIDWM was slower than the reference FM sample. For the SAF structure, a systematic deterioration of the CIDWM occurs as soon as an lr DL is inserted (see Fig. 2f). However, it is worth noting that J _th drops smoothly for Ir layers with thicknesses up to 0.5 nm but increases dramatically for thicker layers.

To directly compare the influence of different dusting layers on the performance of CIDWM, the threshold current density J _th and the DW velocity are plotted at a current density of ~1.2 x 10 ⁸ A/cm ² , as a function of DL thickness t _DL in Fig. 3a-3d for both the FM and SAF structures. The dramatic role of the DL is readily seen and is most pronounced for Pd and Rh DLs. In these cases, the CIDWM efficiency increases substantially for small foL, reaching a maximum at ~0.2 and ~0.1 nm respectively. For lr dusting layers, however, the CIDWM efficiency increases slightly only for the 0.1 nm case and then monotonically decreases until zero propagation as t _DL is increased (Fig. 3a and 3b).

To distinguish CIDWM from DW creep that occurs even at tiny current densities, a threshold current density Jt _h was defined as the current density above which the chiral DW velocity exceeds 5 m/s. Generally, a decrease in J _th was found for racetracks with DLs except for the Ir case. It is worth noting that the spacing between adjacent fee (111) planes of the dusting layers is ~0,22 nm which corresponds to the middle of the colored region in Fig. 3. A plausible argument can thus be made that the efficiency of the chiral DW motion is maximized by inserting one monolayer thick 4 d metal (Pd and Rh) DL. Specifically, the most promising case is the SAF structure with a 0.2 nm thick Pd DL layer: a substantial decrease of ~70% in J _th and a ~350% increase in CIDWM velocity compared to the reference sample was observed at a typical current density of 1.2x10 ¹² A/m ² , exhibiting very fast DW speeds of up to ~1000 m/s.

Magnetic field dependence of DW velocity

The CIDWM is derived from a chiral spin torque, in which the chirality of the DWs in both the FM and SAF structures is stabilized by an interfacial DM! arising from the HM layers with strong spin-orbit coupling. The DW velocity thus depends sensitively on magnetic fields applied along the racetrack: external longitudinal magnetic fields Hx add or subtract from the DMI effective fields that stabilize the chiral DWs. The DW velocity was measured as a function of a longitudinal magnetic field. For simplicity, the movement of DWs with domain configurations under positive current are shown in Fig. 4, in which the racetracks incorporating Pd, Rh, and Ir DLs with thicknesses of 0.1, 0.2, 0.3, and 0.5 nm are presented. Data are shown at a fixed current density (1.2 x 10 ⁸ A/cm ² ). In zero external field both DWs move in the same direction as that of the injected current for both the FM and SAF structures, which is consistent with SOT and ECT. For the FM structure presented in Fig. 4a-4d, the DW velocity shows a linear dependence on Hx, which is a typical behavior as previously observed for this reference structure. The slope varies with the materials and thicknesses of the DL and the sign reverses for DWs, that is This behavior can be readily understood within a 1-D DW analytical model. Such behavior has two key characteristics: the slope of the v-Hx curve, and the magnitude of Hx where the DW velocity goes to zero. Thus, by fitting the v-Hx curves, magnetic properties that cannot be directly measured, especially, the interfacial DMI strength D can be estimated.

For the SAF case, a distinct profile of the v-Hx curve is observed, with a symmetric effect of DW chirality on the DW velocity such that The coupled Neel DWs in the SAF structure undergo a more complicated response in the presence of an external field. Several magnetic properties contribute to the detailed shape of the response curves. A 1-D mode! is used to reproduce the experimental results and extract these parameters. Fig. 4b, 4d, 4f, and 4h show that the variation of DW velocity with external field is increased with thickness of the inserted Pd and Rh DLs. The DW velocity, however, is less sensitive to the thickness of the Ir DL, which should correlate with the relatively weak spin hall effect in Ir. Remarkably, as the thickness of the Ir DL is increased beyond 0.5 nm (Fig. 4h), the corresponding v-Hx curves are mirror profiles of those for thinner Ir DL. We attribute this opposite dependence of DW velocity to the reported opposite sign of interfacial DMI at the Co/lr interfaces compared to Co/Pt interfaces.

Magnetic properties

In order to understand the dependence of CIDWM on dusting layer materials and thicknesses, magnetic properties including saturation magnetization Ms, effective perpendicular magnetic anisotropy constant (defined as where K is the perpendicular magnetic anisotropy an d the hard axis anisotropy field), interfacial DMI constant D and the ratio of the remnant magnetization (magnetization at ~ 0 T) to saturation magnetization (magnetization field ~ 1.5 T) (Mr/Ms) in the SAF structures were measured. The dusting layer thickness dependence of these parameters are plotted in Fig. 5a-5c and 5e. A coherent drop in can be observed with increasing thickness of the DL in all cases (Fig. 5c). Ms shows no systematic change with Pd insertion layer thickness, while a monotonic drop for the Ir and Rh DL cases is observed in Fig. 5a. As is well known, the proximity induced magnetic moments (PIM) in the heavy metal contribute considerably to the total Ms of the FM/HM system. Based on the Stoner criterion (Fig. 1b), it would not be surprising if the PIM decreases when Ir and Rh layers are inserted between the Co and Pt layers. On the other hand, we suppose that there may be a considerable PIM in the Pd dusting layer itself since Pd is very close to the Stoner criteria for magnetism.

The dependence of Mt/Ms on t _DL is shown in Fig. 5b. The changes in M _r /M _s ratio with DL insertion are predominantly due to the variation of Ms in the lower sub-layer of the SAF structure, as can be seen for the FM case in Fig. 5a. It has previously been shown that CIDWM in SAF samples is largely derived from a giant ECT that is increased, the more similar are the two sub-layer moments, i.e. M _t /Ms = 0. Note that the reference SAF sample without any DL has been optimized so that Mt/Ms is close to zero. As discussed above, the efficiency of CIDWM is increased with certain DLs even though this causes M _r /M _s to deviate from zero. Thus, even faster CIDWM in the DL engineered racetracks is promised when Mt/Ms is tuned to zero by small modifications to the racetrack structure.

The fastest DW velocity v _lg in the SAF structure is reached with the aid of an external longitudinal field as shown in Fig. 4. The DL thickness dependence of vig is plotted in Fig. 5d. For the Pd and Rh dusting layers, v _lg increases rapidly and then saturates as the DL is thickened. For Ir DLs, however, v _lg remains almost constant with a gradual decrease when the DL thickness is increased. These results suggests that the DLs employed here can be separated into two groups with respect to their response to H _x or, thereby, the DMI field.

In the FM structure, a monotonic drop in the DMI constant D is observed with increasing DL thickness except for a slight increase for the Rh ~ 0.1 nm case (Fig. 5e). It is worth noticing, as the Ir DL thickness is increased, there is a clear sign change of D.

Manufacturing Method and Measurement Methods

Sample preparation and DW velocity measurement

The samples are deposited by magnetron sputtering at room temperature on Si wafers covered with a S1O2 (thermally oxidized Si) layer (~ 30 nm). All these samples are sandwiched between a bottom TaN layer (~ 2 nm) and a capping TaN layer (~ 5 nm) both with high resistivities. The deposition parameters of these materials were calibrated by quartz crystal microbalance and X-ray reflection. The 50 pm x 3 pm racetrack nanowires are fabricated using photolithography and argon ion milling. All the injected current pulses are fixed at a duration of ~10 ns. The DW velocities are determined from Kerr microscopy measurements. The DWs are created in the racetrack nanowires by injecting pulses of current in the presence of external longitudinal magnetic fields.

TEM Specimen Preparation and investigation The cross-sectional TEM specimens were formed by conventional preparation methods. First, the cross-sections were polished mechanically from both sides. Then, they were dimple-grinded from one side and thinned down to electron transparency by polishing with Ar ions at 5 kV from the other side in a Gatan PIPS (precision ion polishing system) system (Gatan, USA, Pleasanton). For HR-TEM/STEM investigations, a FEI TITAN 80-300 electron microscope with a probe corrector (FEI, USA, Hillsboro) was used at an accelerating voltage of 300 kV. The EDX experiments were performed with a Super-X detector system (4 silicon drift detectors placed symmetrically around the sample area inside objective lens (Oxford, UK, Abingdon)) installed on the microscope for faster and better collection efficiency of X-rays. Acquired EDX maps were analyzed and processed by Bruker Esprit software (Bruker, USA, Billerica).

Magnetic property measurements and calculation of DM! constant D

The magnetizations of the sample were measured in a superconducting quantum interference device (SQUID) at room temperature. is measured using vibrating sample magnetometer measurements, in which the magnetization is recorded with the magnetic field along the hard axis of the films. is defined as the field where the total magnetization rotates from out-of-plane to in-plane. The Mr and Ms of the SAF samples are determined at a field of 0 Oe and 15 kOe, respectively, in the out-of-plane M-H curves. The Horn of the FM structure is extracted, according to a 1-D model, at the field where the DW velocity drops to 0 in the linear fitting of the v-Hx curve. The

DW width is calculated from with A, the exchange stiffness, set to be a constant of 1.0 perg/cm. In the FM structure, when the internal DMI effective field (HDMI) is compensated by an external longitudinal field, the Neel wall structure that is stabilized by DMi is no longer sustained. This results in a minimized SOT and therefore a stationary DW. The DMI constant D is calculated from the expression D = μ ₀ M _s ΔH _{DM I} Derivation of vig from 1-D analytical model on steady motion condition

From the 1-D analytical model based on the DW moment describing the DWM of the SAF system, the DW velocity q could be rewritten as the following forms with a steady- state solution as a fixed DW moment:

Where oci is the damping parameter of each sublayer, with i corresponds to L(lower) or (J(upper) layer; is the m agnetization; is the non-adiabatic constant; is the STT-related (spin transfer torque) DW velocity; g is the gyromagnetic ratio; D is the DW width; Hf is the in-plane shape anisotropy field favoring Bloch wall; is the net longitudinal field including H _x applied and DMi effective field; is the angle between inner magnetization direction of the DW in each layer and x-axis; H is the spin Hall effective field in each layer; J _ex is the interlayer exchange coupling constant.

When an exterior longitudinal field is applied to the system, as indicated from the first equation, the velocity will always peak at the field where the SOTs are maximized as ψ _L = 0 or p since the lower layer experienced larger SOT than the upper layer. If one takes a simple assumption as M _L = M _u = M and neglecting the STT-related term, by taking the condition of ψ _L = 0 or π, then the equation sets above could be rewritten into the following form: For the Pd and Rh case, since the long spin-diffusion length in these materials, one could expect the SOT generated from the Pt bottom layer not decaying so much when traveling through the DL, so the v _lg is directly proportional to the DW width and thus shows a similar saturation behavior of DW width with increasing thickness. For the Ir case, the SOT is largely decreased with increasing the DL thickness, together with a slightly modified DW width. Thus, the vi _g has small variations with varying DL thicknesses.