Controlled Individual Skyrmion Nucleation at Artificial Defects Formed by Ion Irradiation

Nanoscale artificial defects in magnetic multilayer films with PMA and DMI, made in a controllable manner using a FIB, offer deterministic skyrmion nucleation. The nucleated skyrmions are stable over a wider (and lower) range of external field values and the nucleation field is tunable with the type of defect fabricated.

based spintronic device. This phenomenon is studied with a range of transmission electron microscopy techniques to probe quantitatively the magnetic behaviour at the defects with applied field and correlate this with the structural impact of the defects.

Introduction
A magnetic skyrmion is a quasiparticle of magnetisation characterised by its non-trivial topology. They appear in systems with a Dzyaloshinski-Moriya interaction (DMI) -which occurs when there is both a lack of inversion symmetry and strong spin-orbit coupling. [1,2] Skyrmions were first observed in bulk B20 materials (such as MnSi and FeGe [3,4] ) and were found to be great candidates for low energy spintronic devices owing to their small size, stability and high mobility under spin-polarised currents. [5,6] Recent research efforts have centred on finding a material system capable of nucleating individual skyrmions at room temperature and low magnetic fields with the aim of realising a skyrmion-based spintronic device.
DMI is also present at the interface between a ferromagnet and a heavy metal -where the interface itself breaks the inversion symmetry and the heavy metal provides a large spin-orbit interaction. [7][8][9] Pt/Co/Ir is a good example of an interface system where the DMI can reach the high value of 2 mJm −2 , here the asymmetric heavy metal interfaces either side of the magnetic material serve to effectively boost the DMI. [11][12][13] This type of interfacial DMI -which enables the use of polycrystalline materials -allows for easily-deposited, tunable DMI systems in which skyrmions have been observed at room temperature. [12,[14][15][16][17][18] One crucial step necessary for realising a skyrmion-based device lies in finding a reliable and controllable method of nucleating individual skyrmions. A multitude of nucleation methods have been proposed in recent years and these can be put into three categories based on: electrical currents, [10,[19][20][21][22][23][24] laser pulses, [25][26][27] and locally applied electric fields. [28,29] Skyrmions can also nucleate at naturally occurring defects in the material; [30,31] however in this case there is little control over the formation of skyrmions due the random location and variable character of natural defects.
In this paper, we propose an original method of nucleating skyrmions using a focused ion beam (FIB) microscope to create nanoscale artificial defects. In Pt/Co based multilayers, ion irradiation is well-documented to both reduce the perpendicular anisotropy and increase the coercivity in a dose-dependent manner. [34][35][36][37][38] These effects have been exploited to engineer or tailor the local behaviour of magnetic systems. [39][40][41] For example a study found that extended circular areas of FIB irradiation (diameter 300 nm to 1 μm) in multilayers of Pt/Co can stabilise anti-skyrmions and Bloch skyrmions. [41] In contrast to the aforementioned study, here we propose point-like FIB-induced defects to nucleate Néel-type skyrmions.

Results
To explore artificial defects as a mechanism for skyrmion nucleation, point-like defects were created on two distinct multilayered Pt/Co based systems. As will be demonstrated shortly, in their 'as-deposited' state, one of these samples has a high magnetic remanence (close to Ms) and supports both skyrmions and Néel-type walls where the other has a low magnetic remanence (close to zero) and supports only Néel walls (in the field-driven regime). Studying two samples with slightly different layer composition and magnetic behaviour provides a useful test of the reproducibility of this nucleation method while also demonstrating the sensitivity of the method to the multilayer structure.
For transmission electron microscopy (TEM) studies the samples were deposited on substrates with an electron transparent Si3N4 window suspended from a thicker Si frame.
Alternating gradient field magnetometry confirms all samples to support magnetisation out of the sample plane, as is promoted by the interface-induced perpendicular magnetic anisotropy (PMA). [42] . Artificial defects were made with a FEI Nova NanoLab 200 scanning electron microscope (SEM) and FIB. The samples were studied using a JEOL ARM200cF TEM, optimised for magnetic imaging, operated at 200kV. [43] In situ magnetising experiments were performed (at room temperature) on samples 1 and 2 using the Fresnel mode of Lorentz TEM to explore the behaviour of the samples with applied field before and after defect creation.
Details of the magnetic textures at the defects in sample 1 were studied with quantitative, high resolution differential phase contrast (DPC) images. For multilayer materials with PMA, DPC imaging is enabled through advanced processing allowed by pixelated detectors [44][45][46] -this work uses the Medipix3 hybrid pixelated detector, with a Merlin readout system, from Quantum Detectors Ltd. We also explore the structural impact of the defects both in planview with bright field (BF) images and in cross-section with high angle annular dark field (HAADF) images taken from sample 1. The different TEM imaging modes will be discussed later in the paper.
In Figure 1, we show snapshots of the behaviour representative of samples 1 and 2 (unmodified by FIB) in an out-of-plane applied magnetic field in a series of Fresnel images.
In the Fresnel mode of Lorentz TEM, magnetic contrast is visible as either bright or dark lines at the position of domain walls. For these samples with PMA and Néel-type walls the sample must be tilted to get contrast from the out of plane domains. [45][46][47] The electron transparent window on sample 2 is significantly buckled, this surface contortion provides local tilt when the sample is notionally 'untilted' with respect to the thicker flat Si frame as detailed previously. [46] For example in Figure 1 and (d) from sample 2 (buckled surface), were acquired with no explicit sample tilt but the images show clear magnetic contrast indicative of local film tilting. In the TEM we perform in situ magnetising experiments using the objective lens of the microscope, which allows application of an out of plane field variable between 10 mT (the remanent field of the lens) and 2 T. Fields below the remanent value must be set individually using specialised hardware external to the TEM.The Fresnel images in Figure 1 show discrete points in the reversal of the magnetisation from saturation in an out of plane positive field to saturation in a negative field.
Without defects, sample 1 supports skyrmions just before saturation, Figure 1 As discussed earlier, FIB irradiation can alter the magnetic and structural properties of magnetic multilayer films. To explore this effect for skyrmion nucleation, FIB defects were made using a FEI Nova NanoLab 200 scanning electron microscope (SEM) and FIB with a 30 keV Ga + beam energy and beam current of 10 pA -giving an ion beam diameter of 10 nm.
This diameter is defined as the probe full width half maximum (FWHM), but the probe has extensive tails over a larger distance leading to FIB defined features larger than the quoted beam diameter. [40] A wide range of defects were made on sample 1 with the geometry shown in Figure 2 The structural impact of the defects on sample 1 has been studied in both plan view, with BF images, and in cross-section, with HAADF images. For the plan view study, spot defects were made on sample 1. Conversely, for the cross-sectional study, line defects (of equivalent dose to the spot defects) were made on the thicker Si frame of sample 1. A crosssectional electron transparent lamella was prepared from this region using a Xe + plasma FIB.
To protect the sample from damage during the fabrication process, the region was first coated with carbonaceous platinum.
A selection of BF and HAADF images from defect sites created with different ion dose are shown in Figure 2. The top row of Figure 2 shows the BF images in which the granular structure of the polycrystalline sample is visible. The bottom row shows the HAADF images in which the discrete layer structure of the sample is visible. HAADF imaging is essentially atomic number Z imaging, where brighter image contrast corresponds to a higher Z material, hence in these images, in the area corresponding to the multilayer, Ir and Pt are bright (Z = 77, 78 respectively) and Co is dark (Z = 27). A defect of dose 1×10 16 ions/cm 2 is shown in the first column of images, Figure 2(b,e). There is little evidence of damage from this defect in the HAADF image, with each layer of the structure resolvable, similarly the centre of the BF image shows some slight grain enlargement -a known effect of ion irradiation on polycrystalline films [51] -but no grain growth out with the normal distribution of sizes. The 5×10 16 ions/cm 2 defect causes visible damage to the layer structure, Figure    These indicate that the contrast from a saturated defect site is symmetric compared to a defect site with a skyrmion. Moreover, the magnetic contrast in these images is generated by the sample tilt (here +20°), hence any magnetic contrast will reverse with tilt, i.e. at -20° black becomes white and vice versa. Hence, these differences in contrast allow defects and skyrmions to be easily distinguished from one another. Also indicated on Figure 3(d) is the defect diameter and the skyrmion diameter. As mentioned above, the mean defect diameter (averaged over multiple defect sites) is 250±30 nm while the mean skyrmion diameter (in zero field) is 300±10 nm. The similarity between the defect size and skyrmion size suggests that the skyrmion size could be linked to the defect size. For defects smaller than the 'natural' skyrmion size we observe a different relationship, but we discuss this in detail later.
To be technologically advantageous the defects should cause a minimal increase to the depinning current required to move a skyrmion from the defect site. Whilst this is not studied here directly, it is known that local changes in anisotropy increase the pinning field, [23] and that higher ion doses cause a larger modification of the anisotropy. [34][35][36][37][38] Hence, we determine the lowest dose required to cause low-field, room-temperature nucleation for samples 1 and 2.
The enlargement of the grains is thought less likely to be problematic as a previous study identifies the most severe pinning for grain sizes which are the same size as the skyrmions [23] and the FIB enlarged grains are still an order of magnitude smaller than the skyrmions. The behaviour of the samples was again monitored as the applied field was lowered from saturation in a positive out of plane field to saturation in a negative field. Non-local to defect sites the magnetic reversal is the same as in the unmodified samples as discussed in relation to Figure 1. However, as highlighted in red, both samples have specific combinations of applied field and defect dose that cause local nucleation and retention of compact individual skyrmions at 100% of defect sites. Table 1 provides details of the 'ideal' artificial defects found to cause 100% skyrmion nucleation in samples 1 and 2.
Critically, for sample 1 the field range of skyrmion stability includes zero applied field. Sample 2 exhibited very similar behaviour with dose and field to sample 1 although, as expected from a low magnetic remanence system, to retain compact skyrmions a bias field was required. This is a particularly interesting result as skyrmions were not observed at room temperature in sample 2 prior to defect creation. Furthermore, the dose required to nucleate skyrmions is an order of magnitude smaller for sample 2 compared to sample 1demonstrating the extreme sensitivity of this nucleation method to the sample structure. As seen in the cross-sectional images presented in Figure 2(f,g), the energy imparted in the sample by the ions causes damage and intermixing of the multilayer structure: alloying the multilayer. Given the different elemental composition of the two multilayer stacks studied, the magnetic properties of the resulting alloy are certainly different. For example, an older study [52] characterises the magnetisation of binary alloys of Co and various transition metals.
It shows that the magnetisation of Co is more sensitive to alloying with Ru than either Ir or Pt -giving a possible explanation for the greater sensitivity of sample 2 to ion dose than sample 1.
From the spot defects on sample 1, the mean skyrmion size in zero field is 170 ± 30 nm at 5×10 16 ions/cm 2 defects and 180 ± 30 nm at 1×10 17 ions/cm 2 defects (details of these measurements are in the supporting information). From the structural imaging in Figure 2, on sample 1 these defects are associated with an area of damage around 100 nm in diameter. This is in stark contrast to the extended 250 nm diameter defects that nucleate 300 nm diameter skyrmions presented in Figure 3.
The defect nucleated skyrmions in sample 2 are around 300 nm in diameter. As no skyrmions were observed in sample 2 without defects there can be no direct comparisonhowever it is useful to consider the large difference in diameter of skyrmions measured from sample 2 compared to sample 1. The images of defects made on sample 2 show no visible non-magnetic phase contrast, like presented in the Fresnel image in Figure 3(c), or obvious grain growth, like shown in the BF images in Figure 2(b-d). Consequently, although not measured, the lateral size of the defects is almost certainly smaller than measured for sample 1 but the skyrmions are larger.
This information leads us to the conclusion that, for defects smaller than the 'natural' skyrmion size, the size of the defect nucleated skyrmions is determined by the sample properties and not the defect itself. The skyrmion size data relating to sample 1 (both unmodified and with artificial defects) and sample 2 is provided in full in the Supporting Information. The combination of perpendicular magnetisation (the samples must be tilted in order to cause any beam deflection) and large sample thickness compared to the active magnetic thickness mean these materials cause only a small beam deflection but generate considerable, undesirable diffraction contrast associated with the crystallites. This diffraction contrast completely masks the magnetic contrast in standard DPC, however, the more advanced processing enabled by pixelated detectors can reduce this contrast and allow for successful quantitative imaging with DPC. [44][45][46] As outlined in existing studies, [45,46] the beam Lorentz deflection angle β from perpendicularly magnetised materials depends on the sample tilt and is proportional to Bs.  Figure 5(a), near to the defect, and is shown next to the DPC image. By fitting a hyperbolic tangent function to this line trace, the deflection due to the domains was determined as 1.7±0.2 µrad (sample was tilted by 24.6°). From this Bs was calculated to be 1.2±0.1 T, assuming a total magnetic thickness of 6 nm resulting from ten 0.6 nm thick Co layers. This is compared to magnetometry measurements from this sample before irradiation which measured Ms as 1.0 ± 0.1 MAm −1equivalent to 1.2 ± 0.1 T. This quantitative analysis suggests that the defect has caused skyrmion nucleation in all layers, and not just in the surface layers most impacted by the irradiation.

Discussion and Conclusion
In this paper, we have demonstrated that nanoscale artificial defects (created with FIB irradiation) can be used to nucleate Néel-type, isolated, single skyrmions at precise locations in polycrystalline magnetic multilayer systems at room temperature in low, even zero, applied magnetic field. We have studied this effect in different multilayer systems and draw two interesting conclusions. Firstly, in samples that are known to support skyrmions, the defects create an additional pocket of skyrmion stability at a lower applied field strength than without artificial defects. In sample 1 as-grown, skyrmions are observed between -50 to -80 mT but with defects skyrmions are also stable between +15 to -5 mT. Secondly, as observed in sample 2, these artificial defects can even stabilise skyrmions in samples with lower DMI strength that naturally support homochiral Néel walls but never stabilise skyrmions on fieldcycling alone.
The mobility of these FIB nucleated skyrmions remains to be studied. The structural imaging of the defects, Figure 2, indicates that nucleation is most successful at defects with partial layer intermixing -undoubtedly this intermixing is associated with local lowering of the perpendicular anisotropy and DMI strength as both originate from the layer interfaces.
Both of these effects will likely increase the depinning field, as will the local reduction in Ms predicted by alloying. [23,52] We note that the size of the skyrmions nucleated appears uninfluenced by the defect itself so long as it is smaller than the inherent skyrmion size. The skyrmion size in a multilayer system is determined by the interplay of various magnetic energy terms controlled by: the strength of the DMI, the anisotropy, the exchange stiffness and the saturation magnetisation of the material. [17] Hence, even though the skyrmions observed in this study are on the order of 100 nm, we expect this nucleation method to successfully nucleate technologically relevant sub-100 nm skyrmions in an optimised material system. To this end we note that it is possible to create smaller FIB defect sites; for example, here we used a 30 keV, 10 pA focused Ga + beam which has a beam diameter of ≈10 nm but a 35 keV, 10 pA focused He + beam has a beam diameter an order of magnitude smaller and can mill sub-10 nm features. [53,54] It is expected that new device technologies are more likely to be utilised if they mould into current fabrication methods. A relevant example is that, to support skyrmions, polycrystalline systems are desirable over single crystal systems as they fit with current deposition technologies. Focused ion beam microscopes are widely used in device fabrication (for example in fabrication of semiconductor devices and disk read/write heads) hence controlled skyrmion nucleation at artificial FIB defects is certainly a promising mechanism for reproducible generation of room-temperature, zero-field skyrmions.

Experimental Section
Sample preparation: Sample 1 was prepared at the University of Leeds by dc magnetron sputtering using a base pressure of 2×10 −8 mbar and an Ar pressure of 6.7 mbar during deposition. Sample 2 was prepared at CNRS/Thalés by dc magnetron sputtering using a base pressure of 8×10 −8 mbar and an Ar pressure during deposition of 2.5×10 −3 mbar. Both samples were deposited on top of Pt buffers to control the perpendicular magnetic anisotropy and are capped with Pt to prevent oxidation. Magnetic characterisation of the samples prior to defect formation was done using SQUID and AGFM. Artificial defects were created using a FEI Nova NanoLab 200 SEM and FIB using a 30 kV Ga + beam energy and 10 pA current.
The patterning method used to fabricate the defects is discussed in detail in the supplementary information. The cross-sectional lamella was fabricated (from an area containing pre-made artificial defects) using a standard procedure on a FEI Helios SEM and FIB using a 30 kV Xe + beam energy before a final polish was performed with a 5 kV beam energy.
TEM imaging: All TEM imaging was performed on a JEOL ARM 200cF equipped with a cold field emission gun and CEOS probe aberration corrector. The HAADF images were collected in 'objective on' mode using spot size 5, a 40 µm condenser 1 aperture (convergence angle 36 mrad) and a 2 cm camera length. These conditions give a probe size of < 0.2 nm and the HAADF images presented in this paper have a sampling pixel size of 0.2 nm. For the Lorentz TEM images microscope was operated in 'objective-off' mode. The Fresnel images are taken with the instrument in TEM mode with a defocus between 5 and 10 mm. The DPC images are taken with the instrument in STEM mode. The DPC dataset was collected using the Medipix3 hybrid pixelated detector with a Merlin readout system. The DPC image shown in this paper was taken using spot size 1, a 20 µm condenser 1 aperture (convergence angle 1 mrad) and a 1500 cm camera length. This gives a probe size of 3 nm and, in the DPC image presented, the sampling pixel size is 3 nm.
Data availability: Data associated with this work is available from the University of Glasgow: Enlighten Data repository under a CC-BY license at [URL tbc].

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.   -d), and high angle annular dark field (HAADF) images of the defects in cross-section (e-g). The gradual erasing of the layer structure with increasing ion dose can be seen in HAADF images (e-g) -at 10 16 ions/cm 2 the layer structure is mostly undisturbed, at 5×10 16 ions/cm 2 there is a shallow 'u' shape with clear intermixing and at 10 17 ions/cm 2 there is slight milling (highlighted by the white dashed lines) and no layer details remaining. The corresponding changes to the grain structure are seen in BF images (b-d). All images taken on sample 1 and the arrows on HAADF images and lines on BF images are to guide the eye to the centre of the defect.    The table of contents entry should be 50−60 words long and should be written in the present tense and impersonal style (i.e., avoid we). The text should be different from the abstract text.
Nanoscale artificial defects in magnetic multilayer films with PMA and DMI, made in a controllable manner using a FIB, offer deterministic skyrmion nucleation. The nucleated skyrmions are stable over a wider (and lower) range of external field values and the nucleation field is tunable with the type of defect fabricated.  Ga + ion beam, with current 10 pA to give a 10 nm probe.
All defects reported in the main paper were produced using procedure in Figure S1(a) (which irradiates a single point), apart from Figure 3 where the procedure in Figure S1(c) produced larger defects which were used to illustrate the contrast from a skyrmions versus the contrast from a defect. We point out the difference in notional dose using preset methods available on the FIB software which for S1(b) and S1(c) result in a higher dose and larger lateral defect as, effectively, the beam is scanned at multiple overlapping positions. We suspect the severity of this effect to be enhanced when creating objects on the order of the beam diameter. This idea is sketched above the SEM images in Figure S1 where the diffuse red dot represents the ion beam and the black dot shows its centre. The difference in opacity of the graphics represents the unintended higher 'effective' dose caused by beam overlap.
Furthermore, this effect could be lessened by reducing the beam overlap parameter; the default value is 50% and intended to deliver constant dose over the pattern despite the Gaussian shape of the beam.
To give context on the magnitude of the discrepancy in delivered dose discussed in this section, using the procedure in S1(a) (with no beam overlap and therefore unambiguous determination of dose) a 70 nm diameter hole is milled with a dose of 5×10 18 ions/cm 2 . This is compared to the 60 and 110 nm diameter holes milled with the procedures in S1(b, c) respectively with a notational ion dose of 1×10 18 ions/cm 2 .

Line pattern linear beam motion
Circle pattern spiral beam motion

S2. Skyrmion size analysis
As mentioned in the main text, the diameter of a Néel skyrmion can be measured quantitatively from Fresnel images by measuring the distance between the maximum of the peak and minimum of the trough of the skyrmion contrast. For Néel skyrmions, the Fresnel image contrast arises from Mz [1][2][3][4] and principally relates to the gradient of the magnetisation,

S3. Interpreting DPC imaged of Néel-type skyrmions
As mentioned in the preceding section, the in-plane magnetisation associated with Néel-type objects (walls and skyrmions) does not contribute to Lorentz microscopy images. [1][2][3][4] Therefore, for simplicity and without lack of accuracy, the following discussion involves only the domain magnetisation Mz. Lorentz microscopy is sensitive to the B ⊥ , the magnetic induction perpendicular to the electron beam trajectory, integrated through the sample thickness. As discussed in the main text (and the references provided above), this means perpendicularly magnetised materials must be tilted to produce any contrast in Lorentz microscopy images. Both the sensitivity to B ⊥ (not M) and the necessity to tilt, leads to elliptical contrast from a circular Néel skyrmion in DPC images.  Figure S3(e), preferentially reducing the DPC contrast perpendicular to the tilt axis again leading to elliptical contrast. Note that sense of ellipticity in Figure S3(e) is orthogonal to the ellipticity caused by contraction. Parallel to the tilt axis, the skyrmion size is undistorted (as highlighted by the yellow dashed line).
The DPC image simulation was designed to accurately represent the experimental conditions. This uses an Mz configuration constructed to represent the skyrmion imaged in Figure 5 of the main paper. This is a circle of diameter 150 nm with a narrow transitional wall (between ±Mz) matched to the width of the profile in Figure 5(b) -and is calculated at a tilt of 24.6°. If one were to ignore the effect of the stray field (i.e. assuming the contrast only arises from the magnetisation in the sample), the difference between the maximum and minimum values in the DPC image (which occur parallel to the tilt axis) would correspond to 2Bsttan(θ) where Bs = µ0Ms. For a tilt of 24.6°, this is 0.92 Bst. In the simulation, where the effects of the stray field are properly accounted for, the difference between the maximum and minimum values in the DPC image is measured as 0.76 Bst. Therefore, in order to extract Bs (and Ms) experimentally we apply a stray-field-compensating scaling factor of (0.76/0.92 = ) 0.83.
In conclusion, three things should be understood: DPC images of perfectly circularly symmetric skyrmions have elliptical contrast; in the direction perpendicular to the tilt axis, the skyrmion diameter can be measured, as can Ms (assuming the sample thickness is known).
The DPC image calculation is of a perfect system, with no sources of contrast other than the smoothly varying magnetisation. The real samples are polycrystalline, therefore DPC images also contain shorter-range electrostatic grain contrast as well as contrast from the skyrmion.
Consequently, the experimental DPC image in Figure 5(a) of the main text has significant background undulations but otherwise matches Figure S3(f) very well.