arxiv:1611.07691v2 [cond-mat.mes-hall] 19 dec 2016x-ray imaging of spin currents and magnetisation...

Topical Review

X-ray imaging of spin currents and magnetisationdynamics at the nanoscale

Stefano Bonetti∗

∗ Department of Physics, Stockholm University, Sweden

E-mail: [email protected]

Abstract. Understanding how spins move in time and space is the aim ofboth fundamental and applied research in modern magnetism. Over the pastthree decades, research in this field has led to technological advances thathave had a major impact on our society, while improving the understandingof the fundamentals of spin physics. However, important questions still remainunanswered, because it is experimentally challenging to directly observe spins andtheir motion with a combined high spatial and temporal resolution. In this article,we present an overview of the recent advances in X-ray microscopy that allowresearchers to directly watch spins move in time and space at the microscopicallyrelevant scales. We discuss scanning X-ray transmission microscopy (STXM) atresonant soft X-ray edges, which is available at most modern synchrotron lightsources. This technique measures magnetic contrast through the X-ray magneticcircular dichroism (XMCD) effect at the resonant absorption edges, while focusingthe X-ray radiation at the nanometre scale, and using the intrinsic pulsed structureof synchrotron-generated X-rays to create time-resolved images of magnetism atthe nanoscale. In particular, we discuss how the presence of spin currents can bedetected by imaging spin accumulation, and how the magnetisation dynamics inthin ferromagnetic films can be directly imaged. We discuss how a direct lookat the phenomena allows for a deeper understanding of the the physics at play,that is not accessible to other, more indirect techniques. Finally, we present anoverview of the exciting opportunities that lie ahead to further understand thefundamentals of novel spin physics, opportunities offered by the appearance ofdiffraction limited storage rings and free electron lasers.

Submitted to: J. Phys.: Condens. Matter

arX

iv:1

611.

0769

1v2

[co

nd-m

at.m

es-h

all]

19

Dec

201

6

X-ray imaging of spin currents and magnetisation dynamics at the nanoscale 2

1. Introduction

A large part of research in modern magnetism dealswith nanomagnetism, i.e. with those magneticphenomena that take place at the nanometre scale.Nanomagnetism is a fascinating field of investigationthat has advanced rapidly in the last three decadesthanks to an almost unique interplay betweenfundamental and applied research. Such an interplayhas led to several breakthroughs that have greatlyinfluenced our understanding of the fundamentalsof magnetism and, in a few cases, even impactedsociety at large. Following the discovery of the giantmagnetoresistance effect [1, 2] and its implementationin all hard disk drives [3, 4, 5, 6, 7], the end ofthe millennium saw the fundamental discovery ofanother phenomenon that can occur in nanometre-sized magnetic objects: the spin transfer torque effect.This effect was predicted by Slonczewski and Berger in1996 [8, 9] and experimentally observed in 1999 [10].Just a few years ago, the first device based on thespin transfer torque was commercialised by EverspinTechnologies [11], and other companies are planningto enter the market [12, 13]. In addition, there havebeen several other key fundamental discoveries that arecurrently challenging our understanding of magnetism,and that have the potential for applications inthe future: i) the observation of ultrafast (sub-ps)magnetisation dynamics driven by femtosecond laserpulses [14] and all-optical switching of magnetisation[15]; ii) the experimental realization of the spin Halleffect in metals [16, 17]; iii) the discovery of topologicalinsulators [18]. In all cases, there is an opportunity forthe implementation of denser, faster and more energyefficient information storage and manipulation.

One of the key concepts in the field of nanomag-netism is spin current, which is the notion that spinangular momentum can flow in an ordered fashion andbe used to manipulate magnetic order without theneed for magnetic fields, which are generally energy-hungry and difficult to implement at the nanoscale.Great progress has been made in understanding howa spin current can be generated, transported and de-tected by electrical means, and, ultimately, used tomanipulate the magnetisation in nanomagnets. How-ever, it is still challenging to directly image the phe-nomena resulting from the flow of a spin current innanostructures comprising magnetic layers, in partic-ular the appearance of spin accumulation and the oc-

currence of magnetisation dynamics. One must use atool that has not only magnetic sensitivity, necessaryto see spins, but that also combines both high spatialand temporal resolution. There are a few techniquesthat possess two of these three characteristics, such asfemtosecond magneto-optics [19] (magnetic sensitivity,fast) and magnetic force microscopy [20] or electron mi-croscopy [21] (magnetic sensitivity, high spatial resolu-tion), but the combination of all three is still challeng-ing. Time-resolved X-ray microscopy at synchrotronlight sources combines all these three characteristicsat once. Synchrotron-generated X-rays have: a wave-length on the order of 1 nm, which allows for focusingto the nanometre scale; a variable polarisation thatenables magnetic contrast through the X-ray magneticcircular dichroism (XMCD) effect at the resonant ab-sorption edges; and an intrinsic pulsed structure thatcan provide the time resolution.

The goal of this Topical Review is to presentthe progress made in scanning transmission X-raymicroscopy (STXM) at synchrotron light sources andto give an outlook over the opportunities ahead. Itis a particularly exciting moment for the researchcommunity since the first diffraction limited storagering (DLSR) started its operation [22], greatlyenhancing the brightness of synchrotron X-rays, andsince free electron lasers (FELs) have been able toproduce fully coherent photons in the soft and hardX-ray regime at the femtosecond time scales [23, 24].Thanks to these new experimental tools, the futureof time-resolved X-ray imaging looks bright. It isreasonable to expect that researchers will be able tomake “movies” of how spins move in time and spacewith ever increasing temporal and spatial resolutions,and sensitivity, which in turn is expected to leadtowards a deeper understanding of the fundamentalsof magnetism.

The structure of this Topical Review is as follows.The fundamentals of spin currents, magnetisationdynamics and X-rays are revised in Section 2. InSection 3, we describe the working principle of aSTXM setup, currently available at most synchrotronlight sources, that allows for spin-sensitive imagingwith time-resolution. In section 4, we review theexperimental efforts in imaging spin currents andspin-current induced switching of magnetisation, whilein Section 5, we review experiments related to theimaging of magnetisation dynamics. An outlook of thefuture challenges in the field of spin physics, as well


as the new opportunities offered by diffraction limitedstorage rings and free electron lasers are discussed inSection 6.

2. Spin currents, magnetisation dynamics andX-rays: Key concepts

The goal of this section is to review the importanttheoretical concepts needed to understand the imagingexperiments presented in Sections 4 and 5. In thefirst two subsections, we review the fundamentalsof the physics of spin transport and magnetisationdynamics. Key notions within the fields of magnetismand magneto-transport, such as magnetisation, spincurrent and the Stoner model, are assumed to befamiliar to the reader. It is also expected that theconcepts of spin accumulation, spin transfer torqueand the Landau–Lifshitz–Gilbert (LLG) equation havebeen encountered before. In the last subsection, webriefly present the theory necessary to understandhow X-rays can “see” spins. Readers interested ina detailed understanding of the interaction betweenpolarised X-rays and matter are directed towards amore comprehensive text, such as Ref. [25].

2.1. Spin currents: From magnetoresistance to spintransfer torque

The concept of spin current is key to modernmagnetism. The simplest way of picturing a spincurrent is to imagine a flow of electrons in a conductorwith all their spins pointing in a well-defined direction.In this case, the spin current, which is the flow of netspin angular momentum, is collinear and proportionalto the charge current. In other words, each electroncarries its ~/2 of angular momentum along with it.Indeed, this is the situation considered throughoutmost of this Topical Review. However, as we mentionin Section 6, this does not necessarily have to be thecase; the flow of charge and spin can be decoupled, asfor the case of the spin Hall effect.

The most practical way to create a spin currentis to let a spin-unpolarised (or simply “unpolarised”)charge current flow through a ferromagnetic metal.The electrons exiting the ferromagnet have their spinscollinear with the magnetisation of the ferromagnetand, in turn, also carry a net flow of angularmomentum, as shown schematically in Fig. 1. Westress the use of the word collinear rather than parallelor antiparallel, because the sign of the spin polarisationin transport experiments depends on the density ofstates at the Fermi level (qualitatively different forFe as compared to Co and Ni for instance) and onthe sd scattering. The sign of the spin polarisationcan be measured experimentally using spin-polarisedphotoemission [26, 27] or transport [28] experiments.

Figure 1. Schematic of the current flow in a non-magnetic(NM) metal and in a ferromagnetic (FM) one. The current inthe non-magnetic metal has no net spin polarisation, while theflow of charge in a ferromagnet is associated with a flow of spinangular momentum. The spin polarisation is collinear to themagnetisation of the ferromagnet, indicated by the white arrow.Note that the current density J and the flow of electrons haveopposite signs.

Theoretically, it can be determined only by performingaccurate calculations on the specific system of interest.Such considerations are well beyond the scope of thisarticle, and we direct the interested reader towardsa more comprehensive discussion on the topic, suchas Ref. [25]. Since we aim to provide a qualitativedescription of the phenomenon, we will keep it simpleand draw the majority spins and magnetisation parallelto each other with the understanding that in somesystems, the situation could be the opposite.‡

The spin-dependent scattering of the conductionelectrons is the key mechanism that creates spinpolarisation, which can be understood in terms ofthe Stoner model [25]. The band structure of a 3dmetallic ferromagnet such as Fe, Co and Ni, can bedescribed in terms of an “itinerant”, hybridized spband and a “localised” d band. The Stoner modelassumes that the d band is split into two spin dbands shifted in energy (∼ 1 eV) due to the exchangeinteraction. Since the Fermi energy in a material isfixed, the material contains more spins of the band thatis shifted lower in energy, which are named majorityspin states. Conversely, the spins in the band shiftedhigher in energy are called minority spin states. Sincethe total number of states in a given band is fixed (10states/atom for d bands), there are more empty states(i.e. holes) available in the minority band. What thismeans is that the itinerant minority electrons in the sp

‡ There is another important note on the convention used formajority and minority spins. For visual clarity, majority spinstates and magnetisation in a ferromagnet are often drawnparallel to each other. However, the magnetisation is definedas M = m/V , where m is the magnetic moment and V isthe volume of the sample. The magnetic moment is defined asm = −gµBs, which means that spins and magnetisation pointin opposite directions, or that the magnetisation is aligned withthe minority spins. This discussion is left out of the main text toavoid another sign flip to keep track of. None of the importantqualitative discussions presented here depend on such sign.


band have a higher probability to scatter in an emptyd state than the majority electrons, since electron-electron scattering events are spin-conserving and,thus, spin-flip events are forbidden. In other words,the flow of unpolarised electrons into the ferromagnetresults in the minority spin component experiencinga greater resistance in moving through the material,while the majority spins can flow less perturbed. Thenet effect is that, at the exit of the ferromagnet,the original unpolarised current (i.e. made up ofan equal number of majority and minority spins) isspin polarised. The quantization axis for the spins isgiven by the direction of the magnetisation M in thematerial, which can be set by an external magneticfield. Another way of looking at this, according to Mott[29], is that the conductivity in a ferromagnet can bemodeled as an electric circuit made up of two resistorsconnected in parallel, one for each spin channel. Thetotal conductivity is then computed as for any electriccircuit as the sum of the inverse resistivity of the twospin channels.

So far, we have discussed electrons entering andexiting a ferromagnet in an abstract fashion. Inpractice, this is often achieved by placing a thin filmferromagnet in contact with non-magnetic conductors,such as Cu, and use these conducting layers to injectand detect electrons. The technology for depositinglayers of thin metallic films is so well developedthat well-defined interfaces can be assumed in thedescription of most phenomena.

An unpolarised current in a non-magnetic metal(NM) can be thought to be made up of an equal amountof majority and minority spins. When an unpolarisedcurrent flows from a non-magnetic material into aferromagnet, the majority and minority spins havedifferent scattering probabilities. Spins that areparallel to the spin polarisation in the ferromagnethave a higher probability of being transmitted throughthe ferromagnet, while antiparallel spins have a higherprobability of being reflected from the interface. Thisis the so-called spin filter effect. This creates anapparently contradictory situation. Far away from theinterface, in the non-magnetic metal, there must be anequal number of parallel and antiparallel spins withrespect to the spin polarisation in the ferromagnet.However, the net flow of reflected antiparallel spinsseems to imply their buildup in the non-magneticmetal. In reality, at relatively long distances from theinterface (∼ 100 nm for Cu), spins diffuse and electronsbecome effectively unpolarised. However, closer tothe interface, there is an accumulation of antiparallelspins, which is only caused by the spin-dependenttransmissivity and reflectivity of the interface. Thiseffect is called spin accumulation, and the lengthover which it decays is called the spin diffusion

Figure 2. Schematic of a spin-transfer torque induced switchingmechanism in an asymmetric spin valve with lateral dimensionsof the order of 100 nm. The thicker ferromagnet acts asa polariser and a “fixed” reference layer, while the thinnerferromagnetic layer is “free” to rotate under the effect of a torque.A large unpolarised current density J greater than a critical valueJc, switching is sent through the spin valve, where the two layers

have antiparallel magnetisation. The “fixed” layer polarises theelectrons, which can exert a net torque on the “free” magneticlayer and align it parallel to the “fixed” layer. Reversing thecurrent causes the switching from the parallel to the antiparallelconfiguration to occur.

length. Naturally, spin accumulation also occurs whenelectrons leave a ferromagnet and are injected into anon-magnet. In this case, the spins parallel to the spinpolarisation accumulate at the interface with the non-magnetic metal over the same spin diffusion length. Inmany interfaces of interest, such as Cu/Co or Cu/Ni[30], the spin polarisation in the ferromagnet is parallelto the majority spins.

We have thus far neglected the conservation ofangular momentum in the discussion. However, whencreating a spin-polarised current, the spin angularmomentum cannot be created out of nowhere. Thesource of angular momentum is, of course, theferromagnet itself, which implies that, for the angularmomentum to be conserved, the ferromagnet has todepolarise. This is correct, although it is important toconsider the relevant limits.

For current densities below 106 A/cm2, thedepolarisation of the ferromagnet is negligible, and ittakes place at a slow enough rate that the Gilbertdamping in the ferromagnet manages to compensateit. The net effect is that the current becomes spin-polarised with no significant change in the ferromagnet.The situation is more interesting when the currentdensity is greater than 106 A/cm2 and it is spin-polarised. When this net flow of spin angularmomentum is sent through a ferromagnet, it can exerta significant torque on the magnetisation, which iscalled the spin transfer torque effect, as predicted bySlonczewski and Berger two decades ago [8, 9].§ In his

§ One possible and interesting way to estimate the currentdensity needed to observe a measurable torque is as follows.For the spin-polarised current to affect the magnetisation in the


seminal paper [8], Slonczewski predicted that the spintransfer torque could be large enough to be used toswitch the magnetisation of a nanosized ferromagnetwithout the need of magnetic fields. The schematic ofa device based on spin transfer torque is illustrated inFig. 2.

In order for the spin transfer torque to bedominant over other current-induced effects, such asheating and Oersted fields, the current needs to flowin small regions. Typically, nanostructures of ∼ 100nm in lateral size or smaller are necessary, in whichcurrents in the mA range are sufficient to achievethe required current densities, while limiting the heatdeposited in the sample.

2.2. Magnetisation dynamics: From ferromagneticresonance to current-induced spin waves

Ferromagnetic resonance (FMR) is the uniform pre-cession of spins that can be observed in ferromagnetssubjected to oscillating magnetic fields, typically in theGHz range. The resonance is present because the mag-netization obey the LLG equation of motion:

dm

dt= −γm×Heff + α

dm

dt×m, (1)

where γ = g~/µB ≈ 28 GHz/T is the gyromagneticratio and α is the Gilbert damping. The resonancefrequency depends on the magnitude and direction ofthe magnetisation and the applied (static) magneticfield, the shape of the ferromagnet and its internalanisotropies. Typically, one includes all these differentparameters to calculate the total energy in theferromagnet and then computes the effective magneticfield Heff = δE/δM .

Measurements of the FMR are extremely usefulfor the characterisation of thin magnetic films. Inparticular, key parameters, such as the saturationmagnetisation Ms and the Gilbert damping α canbe retrieved in a relatively straightforward way.Conventional techniques measure FMR by exciting

ferromagnet, the number of the flowing spins has to be roughlythe same number as the spins in the ferromagnet. However,these spins cannot take an arbitrarily large amount of time toflow; they have to exert a torque on the magnetisation beforethe Gilbert damping torque restores the equilibrium condition.Assuming a ferromagnet with 1 µB/atom and a density ∼1023

atoms/cm3, and a Gilbert damping rate Γ = αν0 ≈ 100 MHz(reasonable for a Gilbert damping α ≈ 0.01 and a ferromagneticresonance frequency ν0 ≈ 10 GHz), the net flow of spins into thevolume of the ferromagnet needs to be at least 1023 spins/cm3

every 10 ns (1/100 MHz). Assuming a perfectly polarisedspin current (i.e. 1 µB/electron), this can be written as 1012

A/cm3. Since we are dealing with thin films with thicknesses∼10 nm or 10−6 cm, this corresponds to a current densityof 106 A/cm2. This assumes perfectly effective spin transfertorque; the first devices operated with 107-108 A/cm2. However,in recent commercial devices thinner (∼ 1 nm) layers andoptimised structures have been implemented, current densitiesof 106 A/cm2 are sufficient to observe spin torque switching.

Figure 3. Schematic of the excitation of spin waves due tothe spin transfer torque in an asymmetric spin valve. Here,only the “fixed” layer is patterned to a lateral dimension ∼100 nm, while the “free” layer is extended over micrometrelengths. A large unpolarised current density J greater than acritical value Jc, precession is sent through the structure (with

the electrons flowing from right to left), where the two layershave almost parallel magnetisation. The thin “free” layer onlypartly polarises the electrons flowing through it, which are thenfiltered by the “fixed” reference layer. The transmitted electronshave spin “up” polarisation, while the electron reflected at theFM/NM interface have spin “down” polarisation. It is these spin“down” electrons that flow towards the “free” layer and tend todestabilise the magnetisation. With the proper combination ofapplied magnetic field and current, a magnetisation precessioncan be sustained.

the magnetisation of a ferromagnet with an ACmagnetic field and measuring the electromagneticpower absorbed. The resonance condition is foundeither by sweeping the frequency of the AC magneticfield and keeping the larger DC-applied magnetic fieldconstant, or by sweeping the magnitude of the DC fieldand keeping the frequency constant.

As the amplitude of the driving AC magneticfield is increased, however, the response of themagnetisation saturates well below the geometricallimit. This is due to the occurrence of the so-called Suhl instabilities [31], or the coupling of theuniform FMR mode to the non-uniform modes, whichare degenerate in frequency [32]. In spin transfertorque experiments, the nanoscale confinement ofthe excitation, which is necessary to achieve highenough current densities, largely modifies the spin wavespectrum, allowing for the uniform precession mode toreach and overcome geometrical saturation [33].

Slonczewski showed that Eq. (1) can be rewrittento include an extra term:

dm

dt= − γm×Heff + α

dm

dt×m


+ η(θ)µB

e

J

tm× (m×M) , (2)

where J is the current density, t is the film thickness,η(θ) measure the efficiency of the spin torque transfer,and M is the normalised magnetisation of the “fixed”layer that defines the spin-polarisation of the current.Without going into much detail of Eq. (2), referred toas the Landau–Lifshitz–Gilbert–Slonczewski (LLG +S) equation, one can identify a few of its qualitativefeatures. First, the new term vanishes when themagnetisation of “free” and “fixed” layer are collinear,and it is maximised when the magnetisations areorthogonal to each other. Second, the Slonczewskiterm comprises a current density term, which meansthat its sign can be reversed by reversing the currentpolarity. Third, it can be shown that, in theappropriate limit and with the current flowing withthe correct polarity, the Slonczewski term can cancelout the Gilbert damping term. The net effect isthat the right-hand side of Eq. (2) reduces to itsfirst term, i.e. the magnetisation keeps precessingaround the effective field as long as the current isapplied. This cancellation is not achieved exactly atevery instant of time, but over one precession periodon average. In terms of nonlinear dynamics, thesolution of the LLG + S equation is a limit cycle[34]. For completeness, it should be noted that inmagnetic tunnel junctions, spin transfer torque canalso give rise to substantial field-like terms in the LLG+ S equation [35], or terms that have the form ofan effective additional magnetic field around whichthe magnetisation precesses. However, these termsare typically one or more orders of magnitude smallerthan the Slonczewski term (current-like) in metallicmutilayers, which are the ones described in this review.

A schematic of the mechanism of excitation ofmagnetisation dynamics due to spin transfer torque ispresented in Fig. 3. All the qualitatively importantfeatures have been included in the diagram and aredescribed in the caption. When the “free” layeris extended outside the region where the current isfocused, typically by patterning the lead or the “fixed”layer to the nanometre scale, the excited magnetisationdynamics is non-uniform; the precession amplitudeis largest below the region of current flow, and itgradually decreases as the distance from that regionincreases. Two types of spin torque dynamics aretypically observed: i) vortex oscillations, typicallypresent in bigger nanocontacts where the relative effectof the Oersted field is larger; ii) spin waves with eitherpropagating or localised character, mostly excitedin smaller structures where the spin transfer torquedominates.

2.3. How can one “see” spins with X-rays? Thefundamental role of spin-orbit coupling for the X-raymagnetic circular dichroism (XMCD) effect.

Before we discuss the details of X-ray microscopy, it isworthwhile to reflect on how it is possible to see some-thing magnetic using photons. Recalling concepts frombasic electromagnetism, at low frequencies, both theelectric and the magnetic part of the electromagneticradiation can interact with matter, depending on themagnitude of the magnetic permeability and of the di-electric permittivity. For ferromagnets, the magneticpermeability can be extremely large at low frequencies;hower, in the 100 GHz – 1 THz region, the permeabilityis 1 for all materials. The typical pictorial explanationoffered is that, as the frequency of the electromagneticfield increases, it is progressively harder for spins to fol-low the oscillation of the magnetic field component. Ina more quantitative way, this relates to the almost fixedmagnitude of the gyromagnetic ratio γ = gµB/~ ≈ 28GHz/T for electrons, where g is the Lande g-factor, µB

is the Bohr magneton and ~ is the Planck constant,i.e. that is determined only by fundamental constants.The fact that there are no magnetic monopoles (i.e.the magnetic moment of an electron is quantized inunits of µB), and that the spin of the electron has awell-defined angular momentum ~/2 that needs to beconserved, is what suppresses the amplitude of the spinprecession at high frequencies. X-rays have frequenciesof 105–106 THz (1–10 keV), and we can therefore safelyexclude a significant interaction between the spins andthe magnetic component of the X-ray radiation.

Therefore, the electric component of the X-rayradiation must be the one that “sees” the spin,although there are no explicit spin-dependent termsin an electric dipole operator. This seems to be aparadox. The solution to the paradox is found bytaking into consideration the spin-orbit coupling, therelativistic interaction that couples the spin of anelectron with its orbital motion. The orbital motiondefines the electronic distribution around an atom, andthis distribution can be probed with electric fields.

It is important to note that, in general, at X-ray frequencies, even the response of the electronsto electric fields is relatively modest, empiricallydescribed by the fact that the refractive index is 1for all materials in the X-ray region. However, inthe particular case of resonant experiments, whereone tunes the X-ray energy to be the same as theenergy difference between two orbitals, the interactionis greatly amplified, and it becomes comparable tothe strength of the interaction for visible radiation.For 3d elements, which include the room temperatureferromagnets Fe, Co and Ni, the important transitionis the one between the 2p core levels and the 3d bands.If one measures the X-ray absorption around these


Figure 4. Pictorial description of the mechanism of theX-ray magnetic circular dichroism (XMCD) effect. (Left)Schematic view of the two-step process that illustrates thedifferent transition probability for two photons with oppositehelicity tuned at the L2 absorption edge. (Right) ExperimentalXMCD spectra of metallic Fe showing the two absorption peaksat the L3 and L2 edges for sample magnetised parallel (blue),antiparallel (orange) or orthogonal (black) compared with the X-ray propagation direction, as represented by the top-right inset.(From Ref. [25], reprinted with permission.)

transitions, two peaks corresponding to the L3 (2p3/2

to 3d) and L2 (2p1/2 to 3d) absorption edges can beobserved, as shown in Fig. 4 for the case of Fe. Itis around these two transitions that the effect of thespin-orbit coupling is strongest, and the XMCD canbe observed.‖

In its simplest phenomenological explanation,XMCD is the contrast in X-ray absorption observedwhen the magnetisation of a sample is switched. Toobserve such contrast, besides being tuned to anabsorption edge, the X-rays need to be circularlypolarised, and there must be a component of themagnetisation vector that is collinear to the X-raypropagation direction. One way to memorise this is tothink to the circularly polarised X-rays as photons witha spin: when the photon spin is parallel to the spinsin the material, one observes enhanced absorption;when it is antiparallel, absorption is reduced. Photonspin or spins in the material can be equivalentlyswitched to observe the same contrast. Accordingly,the baseline absorption is measured in samples with no-net magnetisation (or with magnetisation orthogonalto the X-ray propagation direction), or using withlinearly polarised X-rays, as shown on the right side

‖ Generally, the X-ray spectral range where XMCD is typicallymeasured in 3d transition metals is called the soft X-ray region.It is conventionally assumed to extend from the carbon K-edge(E ≈ 284 eV) to the copper L2-edge (E ≈ 952 eV). TheL3 edges for the three magnetic elements, Fe, Co and Ni, areat EFe,L3

≈ 707 eV, ECo,L3≈ 778 eV and ENi,L3

≈ 852eV. Many soft X-ray beamlines at modern synchrotrons alsoprovide X-rays over a slightly wider spectrum, and reach theM -edges (3d to 4f transitions) of Gd (EGd,M5

≈ 1190 eV)and Tb (ETb,M5

≈ 1241 eV), rare earth elements that becomestrongly magnetic in proximity of 3d transition metals and thatare interesting from a fundamental and applied perspective.

of Fig. 4. Often, this contrast mechanism is allthat one needs to understand the majority of imagingexperiments.

An important point that needs to be made is thatby using the XMCD contrast at the L−edges, onealways probes the atomic, or the localised, magneticmoment. Therefore, strictly speaking, one cannotdetect a spin current with X-rays, but only anaccumulation of spins, or an induced magnetic momentin proximity of an interface. However, since thedetection of a spin accumulation is associated withthe presence of a spin current, the two concepts aresometimes exchanged.

Given that the XMCD probes the atomic moment,how does the contrast takes place at the microscopiclevel? This can be shown using the two-step modeldescribed in Ref. [25]. The first step is the absorptionof the circularly polarised X-ray photon through thecreation of a photoelectron. Energy is conserved byadding an electron to an empty valence state, althoughthe circular polarisation of the photon also requiresconservation of angular momentum. In general, thisis satisfied by the photoelectron carrying the orbitalangular momentum. In the presence of the spin-orbit coupling in the core levels, the photon angularmomentum is also transferred to the spin of thephotoelectron, which becomes spin-polarised. Thequantization axis is given by the direction of the photonspin, and the sign of the spin polarisation is given bythe sign (±~) of the photon spin and by the sign of thespin orbit coupling, which is opposite for the 2p3/2 and2p1/2 levels.

The second step in the model is the detectionof the spin-polarised photoelectron in the 3d band.As we saw before, in a ferromagnet, the 3d band isdivided into two bands for the majority and minorityspins. Those are split by the exchange interaction,which creates an unequal population of spin-up andspin-down electrons and, conversely, of available stateswith opposite spins. Since the probability of atransition depends on the availability of empty statesfor a photoelectron, photoelectrons with opposite spin-polarisation are absorbed at different rates. In atypical imaging experiment where the energy and thehelicity of the X-rays are fixed (i.e. the number andthe spin-polarisation of the excited photoelectrons arefixed), the relative orientation of the spins in thesample, which defines the spin quantization axis andtherefore the “efficiency” of the 3d bands as detectors,determines the overall absorbed X-ray intensity. Thisis the observed XMCD contrast.


Figure 5. Schematic of a STXM setup available at modern synchrotron light sources. The X-ray photons are produced by thedeflection of ∼ 50 ps long electron bunches separated by a few nanoseconds that travel around in the storage ring. Typical operationfrequencies of the storage rings are in the hundreds of MHz, and revolution frequencies are in the ∼ 1 MHz range. Timing signalsare available at the beamlines requiring them.

3. Scanning transmission X-ray microscopy atsynchrotron light sources

A scanning transmission X-ray microscope [36, 37, 38]is one of the conceptually simplest and most directmicroscopy techniques. An X-ray beam is focusedand sent through the sample, and the transmitted X-ray light is detected with a photodiode. An image isbuilt by scanning the sample, which is mounted onmovable stages, in the plan perpendicular to the X-raypropagation direction. In other words, the image isconstructed pixel-by-pixel from the intensity recordedby the photodiode. A schematic of the setup is shownin Fig. 5.

Focusing the X-rays cannot be realised withconventional lenses, as the refractive index of allmaterials is very close to 1. Instead, Fresnel zoneplates are used, which achieve focusing by constructiveinterference of the X-ray beam. Zone plates aremade of concentric rings (the “zones”) of opaqueand transparent material, typically gold and siliconnitride, respectively. In order to achieve constructiveinterference, the radii rn of the edges (i.e. theboundaries between the opaque and transparent zones)need to decrease as the distance from the centre of thezone plates increases [39]:

r2n = nfλ+ n2λ2/4, (3)

where n is the zone number (to be counted separatelyfor opaque and transparent zones), λ is the X-ray

wavelength and f is the first-order focus. In the longfocal length limit, only the first term in the right-handside of the equation is necessary. In this limit, themaximum possible resolution ∆l is given by:

∆l = 1.22∆rN , (4)

where ∆rN is the width of the outermost zoneplate. This follows from the Rayleigh criterion ∆l =0.61λ/NA, where the numerical aperture is NA =λ/2∆rN [40]. Standard commercial waveplates havea resolution of about 25 nm, and prototypes providing∼ 10 nm resolution have been demonstrated [41].

Typically, only ∼ 10% of the light is focusedinto the first-order focus by the zone plates, with theremaining X-rays being transmitted unfocused (zero-order radiation). To get rid of this part of X-rays,which would otherwise mask the signal from the focus,a precise pin-hole structure, called an order sortingaperture (OSA), is placed in front of the zone plates.

There are many advantages with this type ofmicroscope. i) No or minimal data processing isrequired to retrieve the data, which means thatan image can often be observed on a computerscreen as it is effectively being measured. ii) It isa photon-in/photon-out technique, which allows forgreat versatility in terms of the sample environment.Electric and magnetic fields of, in principle, arbitrarymagnitude can be applied to the sample, in contrastwith techniques that are based on electron yield. Ultra-high vacuum is not required, although a good vacuum


helps reduce the sample contamination, often in termsof carbon deposition. iii) The use of a point detectoralso allows for fast gating of the signal, which is keyto implementing time-resolved capabilities. Even atthe high operation frequencies of synchrotrons, whichproduce ∼ 50 ps X-ray bursts at ∼ 500 MHz rate,detection of the individual bursts is possible using fastavalanche photodiodes (APDs) and advanced photoncounting methods [42].

There are two main issues that need to beaddressed when using STXM: drift and samplefabrication. Drift is naturally present because thetechnique relies on the mechanical translation of stagesover micrometre distances while requiring ∼10 nmaccuracy. However, the reproducibility of such motionis still not good enough using purely mechanical means.Modern commercial STXM setups implement a laserinterferometer mounted on the stages that allows forhighly reproducible motions [43], which solves thisproblem. However, sample fabrication is still only apartially solved issue. In order for the X-rays to betransmitted to the APD, samples need to be preparedon chips made of suspended SiN membranes of ∼ 100nm thickness and ∼ 100 µm in lateral size. While thosemembranes are commercially available, only a subsetof materials can be grown on them. In particular, itis still a challenge to prepare samples with crystallineorder. Luckily, magnetic thin film samples, usedin both applied and fundamental research, are oftenpolycrystalline, which makes them compatible with thegrowth on SiN membranes.

Transient magnetic signals in nanometre-thickmagnetic films may result in a small variation ofthe total X-ray flux, down to a 10−4 − 10−5 relativesignal change. This poses important challenges inthree domains of the measurements. First, thereliable detection of the X-rays transmitted throughthe sample. When measuring the photons comingfrom individual synchrotron bunches, one often ends upin the single-photon regime, which requires advanceddetection methods [42, 44]. Second, the stablesynchronisation between the signals exciting thesample (typically electrical signals) and the detector[45]. Such synchronisation allows for stroboscopicdetection, which is necessary to avoid washing outthe signal and is key to reducing the averaging timeper image. Third, the correction for all drift thatnaturally occurs when measurement times are of theorder of several hours, as is often the case when one hasto achieve an acceptable signal-to-noise ratio and thesignals are small. Since such drift cannot be completelyremoved in large facilities affected by daily thermalexcursions and tidal forces, one possible solution isto have a normalisation signal that follows the verysame drift. For example, one way to achieve this in

Figure 6. Temporal snapshots of spin torque induced switchingin a magnetic spin valve with perpendicular magnetic anisotropy.The colour scale indicates the component of the magnetisationperpendicular to the film and the image plane. The six imagesare recorded at different time intervals from the arrival of acurrent pulse, indicated below each snapshot. (From Ref. [47],reprinted with permission.)

pump/probe experiments is to excite the sample onlyduring even revolutions of the electron orbit, and to usethe odd revolutions to record the normalising images.This has also the advantage to exactly cancel out thenonlinearities in the detector, which is physically thesame for both the reference and the actual signal. Thisclearly requires the ability to gate both the detectionand the excitation signals at the revolution frequency,typically of the order of 1 MHz, but this can often beachieved by commercial electronics. The details on howto implement this to be able to record STXM imagesup to 10 GHz rates can be found in Ref. [46].

4. Imaging spin currents

In this section, we summarise the recent experimentalefforts that have been successful in directly mappingthe presence of spin currents in magnetic nanostruc-tures. First, we describe how STXM has been used torecord images of spin transfer torque switching eventsin nanomagnets with different anisotropies. Then, wediscuss how the same technique has been successfullyimplemented to directly image spin accumulation innon-magnetic Cu, effectively detecting the flow of aspin current.

4.1. Spin torque induced switching

The observation of current-induced magnetisationswitching in a patterned spin valve [10] was thefirst experimental verification of the correctness ofSlonczewski prediction: that spin transfer torque canbe large enough to cause a reversal of the magnetisationtowards a new equilibrium state. However, suchobservation was indirect: the reversal was detected as


a change in the overall resistance of the device throughthe giant magnetoresistance effect.

Therefore, it was highly desirable to directlyobserve the effect of the spin current on thenanomagnet, and to directly image the switching ofthe magnetisation as it happens. A breakthroughhappened in 2006 when Acremann and co-authorswere able to map the switching of the “free” layerof an in-plane magnetised spin valve [48] using theelectrical pump and X-ray probe detection techniquedescribed above. They made a surprising observation:the Oersted field, which was typically assumed tobe negligible compared with the spin transfer torquein structures at the ∼100 nm scales, turned out tobe of fundamental importance. In fact, the Oerstedfield creates an asymmetric potential well (whenan in-plane field is applied to the sample), aroundwhich switching occurs. A subsequent X-ray studyin combination with micromagnetic simulations couldreveal that the switching occurs through the motionof a magnetic vortex [49]. These observations werekey to understanding the mechanism of magnetisationswitching at the nanoscale, which can be bettercontrolled by playing with the magnetic anisotropiesin the system.

A similarly surprising observation was madein a subsequent work [47] using a sample withperpendicular magnetic anisotropy (PMA), i.e. wherethe easy axis of magnetisation is in the out-of-planedirection with respect to the film plane. In this case,the Oersted field had cylindrical symmetry, as theapplied field had no in-plane component that couldbreak the in-plane symmetry. However, even in thiscase, the spin reversal was not uniform, as shownin Fig. 6. One can clearly see that in the ellipticalnanostructure, the reversal is initiated in its centralpart, and it propagates outwards at a later stage.

The shared message from these two works, ob-tained thanks to the unique capabilities of x-ray mi-croscopy, is that in magnetisation reversal, spatial in-homogeneities are fundamental even at the nanometrescale, and that the often used approximation of single-domain, breaks down dramatically.

4.2. Direct imaging of spin currents in non-magneticmaterials

Having succeeded in imaging spin reversal in aferromagnet, the next challenge was to directly observethe spin current that causes the reversal. As wediscussed earlier, we cannot directly observe the flowof a spin, but we can look for the appearance ofspin accumulation across a ferromagnet/non-magnetinterface. As we explained in Section 2, the build upof spin accumulation is a fingerprint of a spin currentflowing through an interface.

Figure 7. Absorption spectrum of Cu in proximity of theL3 edge (gray line) and the Cu XMCD signal (red squares)across a Co/Cu interface traversed by an electrical current. Thesignal is present only in the transient time when the current isswitched on. The transient spectrum (red line) can be fitted bytwo Gaussians (dashed red). (From Ref. [50], reprinted withpermission.)

Observing the occurrence of spin accumulationis, however, extremely challenging, as the associatedmagnetic moment is extremely tiny. For a nanopillar,where the current flows across the layers, the magneticmoment can be estimated as [25]:

m = −D(EF )[µ↑(0)− µ↓(0)

] 1

dµB

∫ d

0

e−x/Λdx, (5)

where D(EF ) is the density of states at the Fermienergy, µ↑(↓)(0) is the chemical potential at theinterface (d = 0) for the majority (minority) spins, dis the thickness of the layer and Λ is the spin diffusionlength. Calculations using realistic parameters for aCo/Cu interface predict mCu ≈ 9× 10−5 µB/atom.

The first experimental proof that a spin accumu-lation signal was tiny came from the experiment byMosendz et al. [51]. They used a full-field microscopeto image the spins injected from a Co film into a Cuwire in a lateral spin valve. The authors could notobserve a signal, but were able to establish an upperlimit for the signal to be detected. According to theircalculations, spin accumulation in Cu should result ina variation of the XMCD signal of the order of 10−4.

The experimental confirmation of both thetheoretical and experimental predictions happenedonly very recently, when spin accumulation at a Co/Cuinterface could be detected spectroscopically usingSTXM [50]. These measurements are summarisedin Fig. 7. The transient XMCD signal caused bythe flow of current resulted in a relative variation of5 × 10−5 over the background X-ray transmission, for


the photon energies around the inflection point of theabsorption peak. These photon energies are the onesthat induce a transition from the 2p core level to theFermi energy EF , with the broadening of the peak dueto the finite energy resolution of the beamline. Therelative variation of the XMCD signal could then beconverted to a transient magnetisation of ∼ 3 × 10−5

µB per Cu atom at EF , within a factor of 3 of thetheoretical value ¶.

While the Fermi level data were consistent withwell-established theories and transport measurements,the spectroscopic data away from the Fermi levelshowed an unexpected feature. Roughly 1 eV abovethe Fermi level, a second peak could be clearlydistinguished. Since the peak was at a relativelylarge distance from the Fermi energy, it could notbe explained in terms of scattering of conductionelectrons in the bulk, with energies that are withina few meV from the Fermi level. Instead, the peakwas found in correspondence of the Fermi energy ofthe interface, where the Cu and Co energy bandshybridize and where the magnetic moment is reduced.The interpretation given to this observation was thatpart of the spins polarised in the Co layer, whileflowing through the interface, got “stuck” as theytransferred angular momentum (via spin torque) tothe interface, which had lost part of its magnetisationdue to hybridization. The quantitative considerationsmade possible by XMCD showed that a couple ofatomic layers could accumulate as many spins as∼30 nm of bulk Cu. This means that roughly halfof the spins were blocked at the interface. Thisinformation, which is extremely challenging to obtainwith conventional transport measurements, may be keyto better understanding spin-dependent scattering inreal interfaces and designing efficient spin injectors.

Very recently, another experimental study re-ported on the observation of a pure AC spin current ina metallic permalloy/Cu/Cu75Mn25/Cu/Co multilayer[52]. In this work, the excitation of the ferromagneticresonance in the permalloy thin film was used to gen-erate a spin current (via the spin-pumping mechanism[53]) in the Cu75Mn25 layer. The spin current was de-tected as a temporal variation of the XMCD signal atthe Mn L3 edge, and measured at the same frequencyof the ferromagnetic resonance in the permalloy layer.By replacing the first Cu layer with an insulating MgOlayer, the spin current was effectively blocked and novariation in the XMCD signal at the Mn edge was ob-served. The authors also showed that the generatedspin current could be used to transfer a spin torqueon the Co layer, demonstrating the flow of net angularmomentum across the multilayer.

¶ It can be shown that the detected signal corresponds to theXMCD signal that would be generated by ∼ 50 Fe atoms.

5. Imaging magnetisation dynamics

In this section, we discuss recent experimental effortsin imaging magnetisation dynamics at the nanoscale.Interest in using X-rays to look at magnetisationdynamics was first stimulated by investigations onmagnetic multilayers where the dynamics between thelayers is coupled [54, 55, 56, 57, 58, 59, 60, 61, 62,63, 64, 65, 66, 67, 68, 69, 70]. One of the uniquecharacteristics of X-rays is that they allow for element-specific detection of the magnetic signal. This canbe simply achieved by tuning their energy, whichcan therefore allow individual layers to be probed.These efforts led to refined schemes for detecting time-resolved XMCD signals, which were then implementedin STXM to image the dynamics.

In recent years, interest in the control ofspin waves in magnetic nanostructures has increaseddramatically, becoming the focus of the research fieldof magnonics [71, 72]. The ultimate goal of magnonicsis to realise circuits where spin waves (i.e. magnons)are used to generate, control and detect information[73] in magnetic nanostructures. In these structures,the confinement introduces additional features tothe spin wave spectrum, such as the appearance ofquantized and edge modes [74, 75]. All these aspectscan be understood only partially using conventionalFMR methods, while imaging techniques are requiredto retrieve the full picture of the physics at play.The work horse in these cases is the technique ofmicrofocused Brillouin light scattering [76, 77], which isable to provide spatial maps of magnetisation dynamicswithout the need of synchrotron radiation. However,the relatively limited spatial resolution of the technique(typically ∼250 nm), as well as the impossibility oflayer-resolved measurements, are sometimes criticalobstacles in nanostructures of fundamental and appliedinterest. Time-resolved X-ray microscopy does nothave such limitations and can be used to image FMRand spin wave modes with extremely high-resolution.

We will discuss imaging experiments of magnetisa-tion dynamics in patterned magnetic structures excitedby AC magnetic fields, in particular vortex dynamics.We will then describe recent studies where the spinwaves induced by spin transfer torque were directlyprobed with time-resolved microscopy.

5.1. Magnetisation dynamics in confined structures:Vortices and skyrmions

The dynamics of spin structures in confined geometriesis distinct from continuous films, as the geometricalconfinement leads to novel spin structures that becauseof their topology exhibit particular dynamics. Ingeneral, shape anisotropy leads to vortex states in discs[80], to geometrically-defined domain walls in wires


Figure 9. (Top row) Time-resolved XMCD images of vortex dynamics triggered by pulsed magnetic fields in a ∼ 50-nm thick,micrometre-sized permalloy (Ni80Fe20) disks. The blue to red contrast measures the amplitude of the magnetisation perpendicularto the thickness of the permalloy and to the image plane. (Middle row) Dynamics of the out-of-plane component of the magnetisationcomputed by micromagnetic simulations using the same parameters of the experiment. (Bottom row) Temporal profile of the magneticfield B (x and y components in red and blue, respectively) that triggers the dynamics. (From Ref. [79], Creative Commons.)

Figure 8. Spatially and temporally resolved images ofmagnetisation dynamics in 1 × 1 µm2 patterned structuresrecorded by time-resolved XMCD. The eight different panelsrepresent eight different phases (45 degrees apart) of oneprecession period of 4 ns (1/250 MHz) of the dynamics. Thewhite to black contrast is a measure of the magnetisation in theplane of the sample along the horizontal direction. The imagescan be analysed to reveal the presence of a magnetic vortex.(From Ref. [78], reprinted with permission.)

[81] and to single skyrmions and geometrically confinedskyrmion lattices in perpendicular anisotropy materials[82].

A particularly interesting spin structure is thevortex. Pioneering experiments in imaging vortexdynamics have been conducted by Stoll and co-authors[83], Choe et al. [84] and Puzic et al. [78] usingfull field X-ray microscopy, the latter experiment being

summarised in Fig. 8. The authors applied a 250 MHzrf field to a 1 × 1 µm2 Ni80Fe20 (permalloy) structurethat was magnetised in-plane. The sample was tiltedwith respect to the plane perpendicular to the X-rayphotons so that there was a finite component of themagnetisation along the X-ray propagation direction,which was necessary to detect an XMCD contrast.Such contrast is presented as a gray scale variation foreight different phases of the precession. The resolutionof the image was not explicitly measured in that work,but it is clearly well below 100 nm, and likely limited bythe sharpness of the magnetic features. By analysingthese spatially and temporally resolved maps, theauthors could explain the features of the images as aconsequence of the gyration of a magnetic vortex core.

After those pioneering imaging experiments, otherinvestigations have been performed using a similarexperimental approach. An important success intime-resolved X-ray imagining with STXM, whichdemonstrated the power of the technique, was thediscovery that a vortex core reverses through theexcitation of the vortex gyrotropic mode and throughthe creation of a vortex-antivortex pair [85]. Theintrinsic elemental selectivity of X-rays was thenused to study the interaction between vortices incoupled magnetic multilayers [86]. STXM alsoallowed researchers to observe the suppression andnucleation of a vortex in proximity of artificially


induced defects [87]. Subsequent time-resolved X-rayimaging experiments have been used to demonstratethat the polarity of a vortex core can selectively beswitched with rotating magnetic fields of oppositehelicity [88], and to image the oscillation of domainwalls [89], as well as stochastic domain-wall depinningin nanowires [90]. Similar studies were extendedtowards the understanding of the dynamics of coupledstructures, and the role of the shape of the magneticstructures in the coupling [91, 92]. In the limit of large-amplitude driving rf magnetic fields, it was also shownthat it is possible to observe non-equilibrium dynamicconfigurations of Landau states in a square permalloydisk that contains a vortex [93]. Those excitationswere identified as the spatial-domain equivalent of thefrequency-domain Suhl instabilities.

Recent studies have been able to map the GHzmagnetic susceptibility in micrometre-sized structures[94], identify a correlation between the velocities ofdomain walls and the transformations of the spinstructure [95] and observe the synchronous motionof multiple domain walls in nanowires [96] and thestochastic formation of vortices [97]. It has also beenshown that the polarisation of a vortex core in apermalloy film can be switched when crossing a domainwall of an adjacent magnetic film with out-of-planeanisotropy in a maze domain state [98]. The possibilityof imaging the dynamics of the vortex domain wallnucleation [99] and the wall propagation dynamics inasymmetric rings [100] also show the importance ofthe local geometry that governs the spin structuredynamics and can be used to tailor it.

Further improvement to the detection sensitivityand the compatibility of STXM with high-frequencyelectronics led to the achievement of spectacular time-resolved imaging capabilities, as illustrated by the workof Kammerer et al.; in this study, they could observethe details of the mechanism of switching of a magneticvortex caused by the formation of a vortex–antivortexpair and subsequent vortex–antivortex annihilation[101]. In follow-up works, the same experimental teamwas able to observe further details of the interactionsbetween spin waves and vortices [102, 103, 104, 79].An example of the images that can be recorded isshown in Fig. 9. The top row of the figure showsthe experimental images, while the middle row themicromagnetic simulations. The agreement betweenthe two series of images is outstanding. A recent anddetailed review on the topic of vortex dynamics imagedby time-resolved XMCD can be found in Ref. [105].

In more recent works, it was demonstrated thatthe vortex core dynamics in antiferromagneticallycoupled layers can be used to create spin waveswith a wavelength that can be tuned linearly withthe frequency of the driving signal [106], and that

collective modes can be observed in arrays of permalloydisks containing magnetic vortices forming a so-called“vortex crystal” [107].

Vortex dynamics can be excited not only byexternal rf magnetic fields, but also by spin transfertorque that can drive large-amplitude, nonlinear vortexdynamics in confined structures. Even in this case,time-resolved XMCD microscopy can be used toimage such dynamics directly. The first experimentalobservation of vortex gyration that could be partiallyexplained by spin transfer torque was reported fora permalloy square with lateral contacts feeding analternating current to the magnetic structure at afrequency of about 60 MHz [108]. In 2011, Yu et al.managed to capture, for the first time, the dynamics ofa magnetic vortex in a nanopillar driven exclusively bythe spin torque provided by the direct current throughthe nanopillar [109]. (An alternate current, two ordersof magnitude smaller than the direct current, was usedonly to phase lock the excitation with the synchrotronfrequency.) By achieving a resolution of 30 nm, theauthors could record a movie of the vortex dynamicsand realise that such dynamics was substantially morecomplicated than usually assumed. In order to performsuch an experiment, a driving signal at a frequencyas high as 1 GHz was required, and the authorsfurther developed the dedicated pump-probe schemeconnected to the STXM described above [42, 45].

Finally, beyond the formation of vortices thatprimarily occurs in soft magnetic materials as aresult of dipolar interactions, qualitatively differentspin structures can form in high anisotropy materials.These structures, such as skyrmions that can bestabilised by additional chiral exchange interactions,can also be imaged using X-ray microscopy [82].

5.2. Spin waves in spin torque oscillators

Since the early works on spin torque induceddynamics, researchers have wondered about thespatial characteristics of the spin waves emittedby a nanocontact into extended magnetic films.Slonczewski originally predicted that an out-of-planemagnetised thin film would emit spin waves witha propagating character, with a wavelength λ =2π(Rc/1.2), where Rc is the radius of the nanocontact[111]. A few years later, Slavin and Tyberkevich [112]predicted that for an in-plane magnetised nanocontact,a localised soliton, so-called spin-wave bullet, wouldmore favourably form underneath the nanocontact.Micromagnetic simulations were able to prove theexistence of both modes, but experimental verificationswere still lacking [113]. A few years later, a combinedexperimental/micromagnetic study [114] demonstratedthat an out-of-plane magnetised contact generatespropagating waves until the equilibrium magnetisation


Figure 10. Time-resolved XMCD images of spin wave dynamics excited in an extended permalloy layer by a spin-polarised currentinjected through a nanocontact (black ellipse). The six images are snapshots at six different phases of the oscillation with a precessionperiod of ∼ 160 ps. The film is in-plane magnetised when no current is applied, and the presence of a contrast (blue to red) inthe images is proportional to the out-of-plane component of the magnetisation that is created by the spin transfer torque. Thecentre of mass of the spin wave is offset by the centre of the nanocontact because of the magnetic potential minimum created bythe superposition of the Oersted field created by the current flowing in the plane perpendicular to the image and an applied staticmagnetic field along the horizontal direction. The thin black lines mark the position of the electrodes, whose topographical featuresare removed to isolate the XMCD contrast. (From Ref. [110], Creative Commons.)

falls below a critical angle θc. Below such angle, bothspin wave modes exist, but they are intermittentlyexcited in time. Subsequent experiments confirmedthis picture [115, 116].

Despite all these efforts, the experiments relied onindirect information retrieved by analysing the spin-wave emission spectra. Two breakthrough experimentswere performed in which microfocused Brillouin LightScattering (µ−BLS) [76, 77] was used to directlylook at the spatial characteristics of the spin wave.At first, Demidov et al. [117] recorded the firstimages of the spin waves emitted by an in-planemagnetised nanocontact, while Madami et al. [118]were able to capture the BLS signal from an out-of-plane magnetised contact. However, the resolution ofthe BLS, diffraction-limited at approximately 250 nm,did not allow for a detailed imaging of the spin wave.Also, the relatively long integration times (∼1 second)needed to detect the spin wave signal, intrinsic of thistype of detection, did not allow for the observation ofthe time-structure of the spin wave.

Only very recently has it been possible to performtime-resolved XMCD experiments with the combinedtemporal and spatial resolution needed to record a

“movie” of the spin wave dynamics at the nanoscalein an in-plane magnetised nanocontact [110]. One ofthe technical steps that had to be implemented was toextend the capabilities of the time-resolved XMCD upto the ∼ 10 GHz rate [46].

The results of this experiment are summarisedin Fig. 10 where six snapshots of the magnetisationdynamics are plotted [110]. The colour scale representsthe amplitude of the magnetisation component parallelto the X-ray propagation direction (orthogonal tothe thin film plane in this case), which is translatedin terms of the out-of-plane cone angle of themagnetisation vector. The sensitivity of the techniqueallows the detection of sub-degree angles, with a noiseon the order of ∼ 0.1 degrees. In this case, the intervalbetween two consecutive snapshots is 27 ps, which isshorter than the ∼ 50 ps FWHM of the X-ray pulses.This means that the measured signal, which is theconvolution of the real signal with a Gaussian function(representing the X-ray pulse), is smaller than theactual signal. The signal-to-noise ratio is, however,large enough to allow for a successful detection of themagnetisation precession.

The most surprising result of this study, enabled


by the combined temporal and spatial resolution, isthat the spin-wave shows a p-like character duringthe precession. This is clearly observed in the plotsat φ = 0 φ = 180 degrees in Fig. 10, wherethe contrast shows a node (i.e. a white region)between two regions where the magnetisation points inopposite directions (red and blue). This fact was notpredicted by previous theories or simulations, whereasa cylindrically symmetric mode, with a s-like character,was expected. However, accurate micromagneticsimulations performed with the same parameters as theexperiment could reproduce the measured data verywell. This study also demonstrated the clear advantagein being able to compare experiments and simulationsnot only in terms of frequencies associated with theprecession of the spin-wave, but also with respect toits spatial profile.

We conclude this section by mentioning thatanother type of spin wave excitation has been predictedin thin films with perpendicular magnetic anisotropy:droplet solitons [119]. The theory predicts that adroplet soliton is a full reversal of the magnetisationbelow the nanocontact where the current is injected.Such droplets have been observed experimentally [120,121], but there are still questions on the details ofthese excitations that indirect transport measurementscannot unambiguously answer. Indeed, work thatlooked at such dynamics directly using magnetic X-ray microscopy showed that the amplitude of theexcitation, directly accessible through normalisation tothe static XMCD contrast, appears to be much smaller(approx. 25 deg) than the predicted full reversal [122].These measurements showed that new, more accuratemodels are needed to describe spin dynamics in thesecases, where the role of thermal fluctuations, disorderin the material and the possible presence of overlyinglower-frequency excitations need to be included.

6. Looking forward: Novel spin physics andnew tools

6.1. Novel spin physics

The spin Hall effect was predicted half a centuryago, and it was observed soon after in semiconductors[123, 124]. However, in metals, it has only recentlybeen observed in experiments[16]. It consists of theappearance of a spin current that is transverse to thecharge current, and it is particularly large in heavymetals with large spin-orbit coupling. In thin films,a charge current flowing in the plane of the materialinduces a spin current towards the interface of thefilm, which can in turn be used to manipulate themagnetisation of an adiacent ferromagnet through aninterfacial torque. Spin Hall induced torques have beendemonstrated with Pt [125], and giant effects observed

with Ta [126] and W [127]. A comprehensive reviewof the spin Hall effect can be found in Ref. [128]. Aneffect related to the spin Hall effect is the flow of a fullypolarised spin current around the edges of topologicalinsulators [18]. In this novel class of materials, aconducting surface state is formed at the boundarybetween the topologically insulating bulk state and thetrivially insulating state of the vacuum (or of anothernon-topological material). The surface state is veryrobust because it is protected by the topology of thebands, and many remarkable properties arise from this.One property is the spin-momentum locking of thecarriers in the surface state, which is the reason for thefull spin polarisation. Recently, it was demonstratedthat the spin current at the interface of a topologicalinsulator can be used to exert a large torque on themagnetisation of an adjacent ferromagnet [129], andthat a magnetically doped topological insulator can beswitched with electric fields [130, 131].

Another possible way to manipulate spin currentswithout the need for magnetic fields is to use magneto-electric multiferroics [132]. In these materials,the ferroelectric and the ferromagnetic orders arecoupled to each other, and the manipulation of theferroelectric polarisation through electric fields affectsthe magnetisation, and vice versa. In noncollinearmagnets, the formation of electric polarisation can beexplained in terms of a spin current [133]. Recently,an ultrafast manipulation of the magnetic orderthrough electric fields has been demonstrated for thenoncollinear magnetic structure [134].

Finally, it has been predicted [135] that ultrafastspin currents can be created using femtosecond laserpulses that induce demagnetisation or magnetisationswitching in thin film ferri- and ferromagnets [14, 15].An observation consistent with the presence of suchultrafast spin currents has recently been made bymeasuring the resonant magnetic X-ray scattering offs X-ray pulses [136], and other experiments seem toconfirm this prediction [137, 138]. However, a directimaging of the effect is still lacking, and competingmechanisms have also been proposed [139, 140]. Thedevelopment of imaging techniques in the ultrafastregime is expected to open up for time-resolved XMCDinvestigations with combined nm and fs resolutionusing lensless holographic techniques [141, 142, 143,144], which will likely contribute towards finding asolution to this outstanding problem.

6.2. New X-ray lightsources: Towards ultrasensitiveand ultrafast imaging

On the instrumentation side, the future for X-ray sci-ence is extremely bright. Two complementary direc-tions are being taken towards developing increasinglybetter X-ray sources: diffraction limited storage rings


(DLSRs) at synchrotron facilities and free electronlasers (FELs).

At the synchrotron facilities, accelerator physicistshave been able to constantly reduce the emittanceand to increase the brightness of the radiation overthe past two decades. However, a quantum leaphappened a decade ago, when a novel design of thestorage ring optics, the multibend achromat, wasintroduced [145]. The new design allows for both anorder of magnitude reduction in the emittance andfor a much more compact synchrotron design, whichalso greatly reduces the construction costs. The firstDLSR synchrotron lightsource based on this design,the MAX IV Laboratory, has just started operationin Lund, Sweden [22]. The same design has beenused to build the SOLARIS synchrotron light sourcein Krakow, Poland [146]. Several other researchlabs in Brazil, France, Japan and United Stateshave ongoing plans for building or upgrading storagerings based on a similar design [147]. The reducedemittance is expected to improve the spatial resolutionbelow 10 nm, and the increased X-ray flux shouldenable more sensitive detection that, in the case ofmagnetism, will push the limits of XMCD microscopytowards the detection of magnetic moments from singleferromagnetic atoms. These capabilities may presentcompletely new opportunities for the study of the novelspin physics at interfaces, as discussed above.

On the FELs side, there is great excitementabout the capability of producing fully coherent andultrafast X-ray photons. Lasing using the principle ofself-amplified spontaneous emission (SASE) [148, 149]was first demonstrated at FLASH in Hamburg in theXUV range [150]. A few years later, groundbreakingSASE X-ray lasing in the soft and hard X-rayregime at the Linac Coherent Light Source (LCLS) atSLAC National Accelerator Laboratory and StanfordUniversity in California, USA was demonstrated [23],which opened the way for the construction of otherFELs around the world. SACLA at the JapanSynchrotron Radiation Research Institute (JASRI) isalready operational [24], while XFEL in Hamburg,Germany [151] and SwissFEL at the Paul ScherrerInstitute (PSI) in Villigen, Switzerland [152] areexpected to start operation in the coming months.The PAL-XFEL at the Pohang Accelerator Laboratoryis being built in South Korea and it is expected tobecome available for users in 2017 [153]. The FERMIElettra free electron laser in Trieste, Italy [154] was thefirst facility to demonstrate full polarisation control ofthe XUV photon, which is of fundamental importancefor the ultrafast imaging of magnetism. Polarisationcontrol at soft X-rays was also recently demonstratedat the LCLS with the implementation of the Deltaundulator [155, 156]. LCLS-II [157], the upgrade of

LCLS to high-repetition rate operation, is expectedto bring another quantum leap in the spectroscopyand imaging capabilities of X-ray lasers using lenslesstechniques.

7. Conclusion

In summary, we reviewed the latest development inmagnetic X-ray microscopy aimed at understandingmagnetism at the nanoscale. STXM at resonantabsorption edges is a particularly suitable tool forthis purpose because of its combined high magneticsensitivity, spatial and temporal resolutions. It isalso an instrument that is commercially available, iscurrently installed at most synchrotron facilities, andhas a performance that is steadily improving. Anotheradvantage is the relatively simple data analysis that isrequired to retrieve images, which greatly speeds updata collection.

We revised the latest experimental results thatused time-resolved XMCD techniques combined withSTXM to image the motion of spins at the nanoscale.We first discussed the detection of spin currents flowingacross metallic nanostructures through the detectionof spin transfer torque switching in ferromagnets andspin accumulation in non-magnetic metals. Then, wepresented the latest developments in the capabilitiesof detecting high-frequency magnetisation dynamics,which have been used to make movies of magneticvortex dynamics and of spin waves excited by spintransfer torque. In all of these experiments, the directobservation of the physics at play brought forth newunderstanding that was not accessible in other indirectmeasurement techniques.

Finally, we presented an overview of the futureopportunities using the new X-ray sources that haverecently been built or that are under construction:diffraction limited storage rings and free electron lasers.It is anticipated that the boost in average or peak X-raybrightness that are provided by these new sources willoffer a whole new range of possibilities for the study ofspin physics with ever increasing sensitivity, speed andspatial resolution. Ultimately, it is expected that thesenew capabilities will allow us to perform experimentswith nanometre and femtosecond resolution. Hence,allowing us to look directly at the temporal andspatial length scales that are relevant to magnetismin condensed matter, with implications for both ourfundamental understanding and our ability to buildspin-based devices.

Acknowledgments

I am grateful to Peter Fischer, Matthias Klaui,Jan Luning, Guido Meier, Gisela Schutz, Hermann


Stoll and Markus Weigand for the useful discussions.Support from the Swedish Research Council grantE0635001 and the Marie Sklodowska Curie Actions,Cofund, Project INCA 600398s is gratefully acknowl-edged.

References

[1] Baibich MN, Broto JM, Fert A, Van Dau FN, PetroffF, Etienne P, et al. Giant magnetoresistance of (001)Fe/(001) Cr magnetic superlattices. Physical reviewletters. 1988;61(21):2472.

[2] Binasch G, Grunberg P, Saurenbach F, Zinn W. Enhancedmagnetoresistance in layered magnetic structures withantiferromagnetic interlayer exchange. Physical reviewB. 1989;39(7):4828.

[3] Parkin SSP, More N, Roche KP. Oscillations in exchangecoupling and magnetoresistance in metallic superlatticestructures: Co/Ru, Co/Cr, and Fe/Cr. PhysicalReview Letters. 1990;64(19):2304.

[4] Parkin SSP, Bhadra R, Roche KP. Oscillatory magneticexchange coupling through thin copper layers. PhysicalReview Letters. 1991;66(16):2152.

[5] Dieny B, Speriosu VS, Metin S, Parkin SS, Gurney BA,Baumgart P, et al. Magnetotransport properties ofmagnetically soft spin-valve structures. Journal ofApplied Physics. 1991;69(8):4774–4779.

[6] Parkin SSP. Origin of enhanced magnetoresistance ofmagnetic multilayers: Spin-dependent scattering frommagnetic interface states. Physical review letters.1993;71(10):1641.

[7] Tsang C, Fontana RE, Lin T, Heim DE, Speriosu VS,Gurney BA, et al. Design, fabrication and testing ofspin-valve read heads for high density recording. IEEETransactions on Magnetics. 1994;30(6):3801–3806.

[8] Slonczewski JC. Current-driven excitation of magneticmultilayers. Journal of Magnetism and MagneticMaterials. 1996;159(1):L1–L7.

[9] Berger L. Emission of spin waves by a magneticmultilayer traversed by a current. Physical Review B.1996;54(13):9353.

[10] Myers EB, Ralph DC, Katine JA, Louie RN, BuhrmanRA. Current-induced switching of domains in magneticmultilayer devices. Science. 1999;285(5429):867–870.

[11] Slaughter JM, Rizzo ND, Janesky J, Whig R, MancoffFB, Houssameddine D, et al. High density ST-MRAMtechnology. In: 2012 International Electron DevicesMeeting; 2012. p. 29.3.1.

[12] Wolf G, Chaves-O’Flynn G, Kent AD, Kardasz B,Watts S, Pinarbasi M. Time resolved transportstudies of magnetization reversal in orthogonal spintransfer magnetic tunnel junction devices. In: SPIENanoScience+ Engineering. International Society forOptics and Photonics; 2014. p. 91671H–91671H.

[13] Nowak J, Robertazzi R, Sun J, Hu G, Park J, Lee J,et al. Voltage and Size Dependence on Write-Error-Rates in STT MRAM down to 11 nm Junction Size.IEEE Magnetics Letters. 2016;7:1–4.

[14] Beaurepaire E, Merle JCC, Daunois A, Bigot JYY.Ultrafast spin dynamics in ferromagnetic nickel.Physical review letters. 1996;76(22):4250.

[15] Stanciu CD, Hansteen F, Kimel AV, Kirilyuk A,Tsukamoto A, Itoh A, et al. All-optical magneticrecording with circularly polarized light. Physicalreview letters. 2007;99(4):047601.

[16] Valenzuela SO, Tinkham M. Direct electronicmeasurement of the spin Hall effect. Nature.2006;442(7099):176–179.

[17] Saitoh E, Ueda M, Miyajima H, Tatara G. Conversion ofspin current into charge current at room temperature:Inverse spin-Hall effect. Applied Physics Letters.2006;88(18):182509.

[18] Zhang H, Liu CXX, Qi XLL, Dai X, Fang Z, Zhang SCC.Topological insulators in Bi2Se3, Bi2Te3 and Sb2Te3with a single Dirac cone on the surface. Nature physics.2009;5(6):438–442.

[19] Kirilyuk A, Kimel AV, Rasing T. Ultrafast opticalmanipulation of magnetic order. Reviews of ModernPhysics. 2010;82(3):2731.

[20] Koblischka MR, Hartmann U. Recent advancesin magnetic force microscopy. Ultramicroscopy.2003;97(1):103–112.

[21] Unguris J, Chung S, Pierce DT. SEMPA Imaging forSpintronics Applications. AIP Conference Proceedings.2007;931(1):472–476.

[22] Eriksson M, van der Veen JF, Quitmann C. Diffraction-limited storage rings–a window to the science of tomor-row. Journal of synchrotron radiation. 2014;21(5):837–842.

[23] Emma P, Akre R, Arthur J, Bionta R, Bostedt C, BozekJ, et al. First lasing and operation of an angstrom-wavelength free-electron laser. Nature Photonics.2010;4(9):641–647.

[24] Ishikawa T, Aoyagi H, Asaka T, Asano Y, Azumi N,Bizen T, et al. A compact X-ray free-electron laseremitting in the sub-angstrom region. nature photonics.2012;6(8):540–544.

[25] Stohr J, Siegmann HC. Magnetism: from fundamentalsto nanoscale dynamics. vol. 152. Springer; 2007.

[26] Busch G, Campagna M, Cotti P, Siegmann HC.Observation of electron polarization in photoemission.Physical Review Letters. 1969;22(12):597.

[27] Banninger U, Busch G, Campagna M, Siegmann HC.Photoelectron spin polarization and ferromagnetism ofcrystalline and amorphous nickel. Physical ReviewLetters. 1970;25(9):585.

[28] Vouille C, Barthelemy A, Mpondo FE, Fert A, SchroederPA, Hsu SY, et al. Microscopic mechanismsof giant magnetoresistance. Physical Review B.1999;60(9):6710.

[29] Mott NF. The electrical conductivity of transition metals.In: Proceedings of the Royal Society of London A:Mathematical, Physical and Engineering Sciences. vol.153. The Royal Society; 1936. p. 699–717.

[30] Stiles MD, Zangwill A. Anatomy of spin-transfer torque.Physical Review B. 2002;66(1):014407.

[31] Suhl H. The theory of ferromagnetic resonance at highsignal powers. Journal of Physics and Chemistry ofSolids. 1957;1(4):209–227.

[32] Silva TJ, Rippard WH. Developments in nano-oscillators based upon spin-transfer point-contactdevices. Journal of Magnetism and Magnetic Materials.2008;320(7):1260–1271.

[33] Bonetti S, Puliafito V, Consolo G, Tiberkevich VS, SlavinAN, Akerman J. Power and linewidth of propagatingand localized modes in nanocontact spin-torque oscilla-tors. Physical Review B. 2012;85(17):174427.

[34] Bertotti G, Serpico C, Mayergoyz ID, Magni A, d’AquinoM, Bonin R. Magnetization switching and microwaveoscillations in nanomagnets driven by spin-polarizedcurrents. Physical review letters. 2005;94(12):127206.

[35] Zhang S, Levy PM, Fert A. Mechanisms of spin-polarizedcurrent-driven magnetization switching. Physicalreview letters. 2002;88(23):236601.

[36] Rarback H, Shu D, Feng SC, Ade H, Kirz J, McNulty I,et al. Scanning x-ray microscope with 75-nm resolution.Review of scientific instruments. 1988;59(1):52–59.

[37] Jacobsen C, Williams S, Anderson E, Browne MT, Buckley


CJ, Kern D, et al. Diffraction-limited imaging ina scanning transmission x-ray microscope. OpticsCommunications. 1991;86(3-4):351–364.

[38] Fischer P, Ohldag H. X-rays and magnetism. Reports onProgress in Physics. 2015;78(9):094501.

[39] Kirz J, Attwood D. X-ray data booklet. X-Ray DataBooklet. 1986;p. 13.

[40] Attwood D. Soft x-rays and extreme ultraviolet radiation:principles and applications. Cambridge universitypress; 2007.

[41] Chao W, Kim J, Rekawa S, Fischer P, Anderson EH.Demonstration of 12 nm resolution Fresnel zone platelens based soft X-ray microscopy. Optics Express.2009;17(20):17669–17677.

[42] Acremann Y, Chembrolu V, Strachan JP, Tyliszczak T,Stohr J. Software defined photon counting system fortime resolved x-ray experiments. Review of ScientificInstruments. 2007;78(1):014702.

[43] Kilcoyne ALD, Tyliszczak T, Steele WF, Fakra S,Hitchcock P, Franck K, et al. Interferometer-controlled scanning transmission X-ray microscopes atthe Advanced Light Source. Journal of synchrotronradiation. 2003;10(2):125–136.

[44] Kaznatcheev K, Bertwistle D, Cheng C, Zohar S, BaileyWE. Synchronous (Lockin) Measurement Techniquesfor Magnetic Contrast Enhancement in STXM. AIPConference Proceedings. 2011 9;1365(1):333–336.

[45] Strachan JP, Chembrolu V, Yu XW, Tyliszczak T,Acremann Y. Synchronized and configurable sourceof electrical pulses for x-ray pump-probe experiments.Review of scientific instruments. 2007;78(5):054703.

[46] Bonetti S, Kukreja R, Chen Z, Spoddig D, Ollefs K,Schoppner C, et al. Microwave soft x-ray microscopyfor nanoscale magnetization dynamics in the 5–10 GHzfrequency range. Review of Scientific Instruments.2015;86(9):093703.

[47] Bernstein DP, Brauer B, Kukreja R, Stohr J, HauetT, Cucchiara J, et al. Nonuniform switching ofthe perpendicular magnetization in a spin-torque-driven magnetic nanopillar. Physical Review B.2011;83(18):180410.

[48] Acremann Y, Strachan JP, Chembrolu V, Andrews SD,Tyliszczak T, Katine JA, et al. Time-resolved imagingof spin transfer switching: beyond the macrospinconcept. Physical review letters. 2006;96(21):217202.

[49] Strachan JP, Chembrolu V, Acremann Y, Yu XW, Tu-lapurkar AA, Tyliszczak T, et al. Direct obser-vation of spin-torque driven magnetization reversalthrough nonuniform modes. Physical review letters.2008;100(24):247201.

[50] Kukreja R, Bonetti S, Chen Z, Backes D, Acremann Y,Katine JA, et al. X-ray detection of transient magneticmoments induced by a spin current in Cu. Physicalreview letters. 2015;115(9):096601.

[51] Mosendz O, Mihajlovic G, Pearson JE, Fischer P, ImMYY, Bader SD, et al. Imaging of lateral spinvalves with soft x-ray microscopy. Physical Review B.2009;80(10):104439.

[52] Li J, Shelford LR, Shafer P, Tan A, Deng JX, Keatley PS,et al. Direct Detection of Pure ac Spin Current by X-Ray Pump-Probe Measurements. Phys Rev Lett. 2016Aug;117:076602.

[53] Tserkovnyak Y, Brataas A, Bauer GE. Enhanced Gilbertdamping in thin ferromagnetic films. Physical ReviewLetters. 2002;88(11):117601.

[54] Bonfim M, Ghiringhelli G, Montaigne F, Pizzini S, BrookesNB, Petroff F, et al. Element-selective nanosecondmagnetization dynamics in magnetic heterostructures.Physical review letters. 2001;86(16):3646.

[55] Vogel J, Kuch W, Bonfim M, Camarero J, Pennec Y, Offi

F, et al. Time-resolved magnetic domain imaging by x-ray photoemission electron microscopy. Applied physicsletters. 2003;82(14):2299–2301.

[56] Bailey WE, Cheng L, Keavney DJ, Kao CCC, VescovoE, Arena DA. Precessional dynamics of elementalmoments in a ferromagnetic alloy. Physical Review B.2004;70(17):172403.

[57] Arena DA, Vescovo E, Kao CCC, Guan Y, BaileyWE. Weakly coupled motion of individual layersin ferromagnetic resonance. Physical Review B.2006;74(6):064409.

[58] Guan Y, Bailey WE, Kao CCC, Vescovo E, Arena DA.Comparison of time-resolved x-ray magnetic circulardichroism measurements in reflection and transmissionfor layer-specific precessional dynamics measurements.Journal of applied physics. 2006;99(8):08J305.

[59] Arena DA, Vescovo E, Kao CCC, Guan Y, BaileyWE. Combined time-resolved x-ray magnetic circulardichroism and ferromagnetic resonance studies ofmagnetic alloys and multilayers. Journal of appliedphysics. 2007;101(9):09C109.

[60] Guan Y, Bailey WEE, Vescovo E, Kao CC, ArenaDAA. Phase and amplitude of element-specific momentprecession in. Journal of Magnetism and MagneticMaterials. 2007 5;312(2):374–378.

[61] Martin T, Woltersdorf G, Stamm C, Durr HA, Mattheis R,Back C, et al. Layer resolved magnetization dynamicsin interlayer exchange coupled Ni81Fe19/Ru/Co90Fe10by time resolved x-ray magnetic circular dichroism.Journal of Applied Physics. 2008;103(7):07B112.

[62] Martin T, Woltersdorf G, Stamm C, Durr HA, MattheisR, Back CH, et al. Layer resolved magnetiza-tion dynamics in coupled magnetic films using time-resolved x-ray magnetic circular dichroism with con-tinuous wave excitation. Journal of Applied Physics.2009;105(7):07D310.

[63] Arena DA, Ding Y, Vescovo E, Zohar S, Guan Y, BaileyWE. A compact apparatus for studies of elementand phase-resolved ferromagnetic resonance. Review ofScientific Instruments. 2009 8;80(8):083903.

[64] Boero G, Rusponi S, Kavich J, Rizzini AL, Piamonteze C,Nolting F, et al.; s. Longitudinal detection of ferromag-netic resonance using x-ray transmission measurements.Rev Sci Instrum. 2009 12;80(12):123902.

[65] Goulon J, Rogalev A, Goujon G, Wilhelm F, Ben YoussefJ, Gros C, et al. X-ray detected magnetic reso-nance: a unique probe of the precession dynamicsof orbital magnetization components. Int J Mol Sci.2011;12(12):8797–835.

[66] Cheng C, Kaznatcheev K, Bailey WE. Stochastic limitsin synchronous imaging of sub-micron magnetizationdynamics using scanning transmission x-ray microscopy.Journal of Applied Physics. 2012;111(7):07E321.

[67] Rogalev A, Goulon J, Goujon G, Wilhelm F, Ogawa I,Idehara T. X-Ray Detected Magnetic Resonance atSub-THz Frequencies Using a High Power GyrotronSource. Journal of Infrared, Millimeter, and TerahertzWaves. 2012 7;33(7):777–793.

[68] Warnicke P, Knut R, Wahlstrom E, Karis O, Bailey WE,Arena DA. Exploring the accessible frequency rangeof phase-resolved ferromagnetic resonance detectedwith x-rays. Journal of Applied Physics. 20131;113(3):033904.

[69] Bailey WE, Cheng C, Knut R, Karis O, Auffret S, ZoharS, et al. Detection of microwave phase variation innanometre-scale magnetic heterostructures. Naturecommunications. 2013;4:2025.

[70] Stenning GBG, Shelford LR, Cavill SA, Hoffmann F,Haertinger M, Hesjedal T, et al. Magnetizationdynamics in an exchange-coupled NiFe/CoFe bilayer


studied by x-ray detected ferromagnetic resonance.New Journal of Physics. 2015;17(1):013019.

[71] Neusser S, Grundler D. Magnonics: spin waves onthe nanoscale. Advanced Materials. 2009;21(28):2927–2932.

[72] Kruglyak VV, Demokritov SO, Grundler D. Magnon-ics. Journal of Physics D: Applied Physics.2010;43(26):264001.

[73] Bonetti S. Nano-contact spin-torque oscillators asmagnonic building blocks. In: Magnonics. Springer;2013. p. 177–187.

[74] Demokritov SO. Spin wave confinement. Pan StanfordPublishing; 2009.

[75] Cheng C, Cao W, Bailey WE. Phase-resolved imaging ofedge-mode spin waves using scanning transmission x-ray microscopy. Journal of Magnetism and MagneticMaterials. 2017;424:12–15.

[76] Demidov VE, Demokritov SO, Hillebrands B, LaufenbergM, Freitas PP. Radiation of spin waves by a singlemicrometer-sized magnetic element. Applied physicsletters. 2004;85(14):2866–2868.

[77] Sebastian T, Schultheiss K, Obry B, Hillebrands B,Schultheiss H. Micro-focused Brillouin light scattering:imaging spin waves at the nanoscale. Frontiers inPhysics. 2015;3:35.

[78] Puzic A, Van Waeyenberge B, Chou KW, Fischer P, StollH, Schutz G, et al. Spatially resolved ferromagneticresonance: Imaging of ferromagnetic eigenmodes.Journal of applied physics. 2005;97(10):10E704.

[79] Noske M, Gangwar A, Stoll H, Kammerer M, SprollM, Dieterle G, et al. Unidirectional sub-100-psmagnetic vortex core reversal. Physical Review B.2014;90(10):104415.

[80] Shinjo T, Okuno T, Hassdorf R, Shigeto K, Ono T.Magnetic vortex core observation in circular dots ofpermalloy. Science. 2000;289(5481):930–932.

[81] Boulle O, Malinowski G, Klaui M. Current-induced do-main wall motion in nanoscale ferromagnetic elements.Materials Science and Engineering: R: Reports. 20119;72(9):159–187.

[82] Woo S, Litzius K, Kruger B, Im MYY, Caretta L,Richter K, et al. Observation of room-temperaturemagnetic skyrmions and their current-driven dynamicsin ultrathin metallic ferromagnets. Nature materials.2016;p. 501506.

[83] Stoll H, Puzic A, Van Waeyenberge B, Fischer P, RaabeJ, Buess M, et al. High-resolution imaging of fastmagnetization dynamics in magnetic nanostructures.Applied Physics Letters. 2004;84(17):3328–3330.

[84] Choe SB, Acremann Y, Scholl A, Bauer A, Doran A, StohrJ, et al. Vortex Core-Driven Magnetization Dynamics.Science. 2004 4;304(5669):420–422.

[85] Van Waeyenberge B, Puzic A, Stoll H, Chou KW,Tyliszczak T, Hertel R, et al. Magnetic vortex corereversal by excitation with short bursts of an alternatingfield. Nature. 2006;444(7118):461–464.

[86] Chou KW, Puzic A, Stoll H, Schutz G, Van WaeyenbergeB, Tyliszczak T, et al. Vortex dynamics in coupledferromagnetic multilayer structures. Journal of appliedphysics. 2006;99(8):08F305.

[87] Kuepper K, Bischoff L, Akhmadaliev C, FassbenderJ, Stoll H, Chou KW, et al. Vortex dynamics inPermalloy disks with artificial defects: Suppressionof the gyrotropic mode. Applied physics letters.2007;90(6):062506.

[88] Curcic M, Van Waeyenberge B, Vansteenkiste A, WeigandM, Sackmann V, Stoll H, et al. Polarizationselective magnetic vortex dynamics and core reversalin rotating magnetic fields. Physical review letters.2008;101(19):197204.

[89] Bocklage L, Kruger B, Eiselt R, Bolte M, FischerP, Meier G. Time-resolved imaging of current-induced domain-wall oscillations. Physical Review B.2008;78(18):180405.

[90] Im MYY, Bocklage L, Fischer P, Meier G. Di-rect observation of stochastic domain-wall depin-ning in magnetic nanowires. Physical review letters.2009;102(14):147204.

[91] Jung H, Lee KSS, Jeong DEE, Choi YSS, Yu YSS, HanDSS, et al. Tunable negligible-loss energy transferbetween dipolar-coupled magnetic disks by stimulatedvortex gyration. Scientific reports. 2011;1.

[92] Vogel A, Kamionka T, Martens M, Drews A, Chou KW,Tyliszczak T, et al. Coupled vortex oscillations inspatially separated permalloy squares. Physical reviewletters. 2011;106(13):137201.

[93] Stevenson SE, Moutafis C, Heldt G, Chopdekar RV,Quitmann C, Heyderman LJ, et al. Dynamicstabilization of nonequilibrium domain configurationsin magnetic squares with high amplitude excitations.Physical Review B. 2013;87(5):054423.

[94] Cheng C, Bailey WE. Sub-micron mapping ofGHz magnetic susceptibility using scanning trans-mission x-ray microscopy. Applied Physics Letters.2012;101(18):182407.

[95] Bisig A, Stark M, Mawass MAA, Moutafis C, Rhensius J,Heidler J, et al. Correlation between spin structureoscillations and domain wall velocities. Naturecommunications. 2013;4:2328.

[96] Kim JSS, Mawass MAA, Bisig A, Kruger B, Reeve RM,Schulz T, et al. Synchronous precessional motion ofmultiple domain walls in a ferromagnetic nanowire byperpendicular field pulses. Nature communications.2014;5:3429.

[97] Im MYY, Lee KSS, Vogel A, Hong JII, Meier G, FischerP. Stochastic formation of magnetic vortex structuresin asymmetric disks triggered by chaotic dynamics.Nature communications. 2014;5.

[98] Wohlhuter P, Bryan MT, Warnicke P, Gliga S, StevensonSE, Heldt G, et al. Nanoscale switch for vortexpolarization mediated by Bloch core formation inmagnetic hybrid systems. Nature communications.2015;6.

[99] Richter K, Krone A, Mawass MAA, Kruger B, WeigandM, Stoll H, et al. Localized domain wall nucleationdynamics in asymmetric ferromagnetic rings revealedby direct time-resolved magnetic imaging. PhysicalReview B. 2016;94(2):024435.

[100] Richter K, Krone A, Mawass MAA, Kruger B, Weigand M,Stoll H, et al. Local Domain-Wall Velocity Engineeringvia Tailored Potential Landscapes in FerromagneticRings. Physical Review Applied. 2016;5(2):024007.

[101] Kammerer M, Weigand M, Curcic M, Noske M, Sproll M,Vansteenkiste A, et al. Magnetic vortex core reversalby excitation of spin waves. Nature communications.2011;2:279.

[102] Kammerer M, Stoll H, Noske M, Sproll M, Weigand M,Illg C, et al. Fast spin-wave-mediated magnetic vortexcore reversal. Physical Review B. 2012;86(13):134426.

[103] Sproll M, Noske M, Bauer H, Kammerer M, Gangwar A,Dieterle G, et al. Low-amplitude magnetic vortex corereversal by non-linear interaction between azimuthalspin waves and the vortex gyromode. Applied PhysicsLetters. 2014;104(1):012409.

[104] Bauer HG, Sproll M, Back CH, Woltersdorf G. Vortexcore reversal due to spin wave interference. Physicalreview letters. 2014;112(7):077201.

[105] Stoll H, Noske M, Weigand M, Richter K, KrugerB, Reeve RM, et al. Imaging spin dynamics onthe nanoscale using X-Ray microscopy. Frontiers in


Physics. 2015;3:26.[106] Wintz S, Tiberkevich V, Weigand M, Raabe J, Lindner J,

Erbe A, et al. Magnetic vortex cores as tunable spin-wave emitters. Nature Nanotechnology. 2016;p. 948953.

[107] Hanze M, Adolff CF, Schulte B, Moller J, WeigandM, Meier G. Collective modes in three-dimensionalmagnonic vortex crystals. Scientific reports. 2016;6.

[108] Bolte M, Meier G, Kruger B, Drews A, Eiselt R, BocklageL, et al. Time-resolved x-ray microscopy of spin-torque-induced magnetic vortex gyration. Physical reviewletters. 2008;100(17):176601.

[109] Yu XW, Pribiag VS, Acremann Y, Tulapurkar AA,Tyliszczak T, Chou KW, et al. Images of a Spin-Torque-Driven Magnetic Nano-Oscillator. Phys RevLett. 2011 4;106(16):167202.

[110] Bonetti S, Kukreja R, Chen Z, Macia F, Hernandez JM,Eklund A, et al. Direct observation and imaging ofa spin-wave soliton with p-like symmetry. NatureCommunications. 2015 11;6.

[111] Slonczewski JC. Excitation of spin waves by an electriccurrent. Journal of Magnetism and Magnetic Materials.1999;195(2):L261–L268.

[112] Slavin A, Tiberkevich V. Spin wave mode excited byspin-polarized current in a magnetic nanocontact is astanding self-localized wave bullet. Physical reviewletters. 2005;95(23):237201.

[113] Consolo G, Azzerboni B, Gerhart G, Melkov GA,Tiberkevich V, Slavin AN. Excitation of self-localizedspin-wave bullets by spin-polarized current in in-planemagnetized magnetic nanocontacts: a micromagneticstudy. Physical Review B. 2007;76(14):144410.

[114] Bonetti S, Tiberkevich V, Consolo G, Finocchio G,Muduli P, Mancoff F, et al. Experimental evidenceof self-localized and propagating spin wave modesin obliquely magnetized current-driven nanocontacts.Physical review letters. 2010;105(21):217204.

[115] Consolo G, Finocchio G, Siracusano G, Bonetti S, EklundA, Akerman J, et al. Non-stationary excitationof two localized spin-wave modes in a nano-contactspin torque oscillator. Journal of Applied Physics.2013;114(15):153906.

[116] Dumas RK, Iacocca E, Bonetti S, Sani SR, Mohseni SM,Eklund A, et al. Spin-wave-mode coexistence on thenanoscale: a consequence of the oersted-field-inducedasymmetric energy landscape. Physical review letters.2013;110(25):257202.

[117] Demidov VE, Urazhdin S, Demokritov SO. Directobservation and mapping of spin waves emittedby spin-torque nano-oscillators. Nature materials.2010;9(12):984–988.

[118] Madami M, Bonetti S, Consolo G, Tacchi S, Carlotti G,Gubbiotti G, et al. Direct observation of a propagatingspin wave induced by spin-transfer torque. Naturenanotechnology. 2011;6(10):635–638.

[119] Hoefer MA, Silva TJ, Keller MW. Theory for a dissipativedroplet soliton excited by a spin torque nanocontact.Physical Review B. 2010;82(5):054432.

[120] Mohseni SM, Sani SR, Persson J, Nguyen TNA, Chung S,Pogoryelov Y, et al. Spin torque–generated magneticdroplet solitons. Science. 2013;339(6125):1295–1298.

[121] Macia F, Backes D, Kent AD. Stable magneticdroplet solitons in spin-transfer nanocontacts. Naturenanotechnology. 2014;9(12):992–996.

[122] Backes D, Macia F, Bonetti S, Kukreja R, Ohldag H,Kent AD. Direct Observation of a Localized MagneticSoliton in a Spin-Transfer Nanocontact. Physicalreview letters. 2015;115(12):127205.

[123] Chazalviel JN, Solomon I. Experimental evidence of theanomalous Hall effect in a nonmagnetic semiconductor.Physical Review Letters. 1972;29(25):1676.

[124] Chazalviel JNN. Spin-dependent Hall effect in semicon-ductors. Physical Review B. 1975;11(10):3918.

[125] Liu L, Lee OJ, Gudmundsen TJ, Ralph DC, Buhrman RA.Current-induced switching of perpendicularly magne-tized magnetic layers using spin torque from the spinHall effect. Physical review letters. 2012;109(9):096602.

[126] Liu L, Pai CFF, Li Y, Tseng HW, Ralph DC, BuhrmanRA. Spin-torque switching with the giant spin Halleffect of tantalum. Science. 2012;336(6081):555–558.

[127] Pai CFF, Liu L, Li Y, Tseng HW, Ralph DC, BuhrmanRA. Spin transfer torque devices utilizing the giantspin Hall effect of tungsten. Applied Physics Letters.2012;101(12):122404.

[128] Sinova J, Valenzuela SO, Wunderlich J, Back CH,Jungwirth T. Spin Hall effects. Reviews of ModernPhysics. 2015;87(4):1213.

[129] Mellnik AR, Lee JS, Richardella A, Grab JL, Mintun PJ,Fischer MH, et al. Spin-transfer torque generated bya topological insulator. Nature. 2014 7;511(7510):449–451.

[130] Fan Y, Upadhyaya P, Kou X, Lang M, Takei S, Wang Z,et al. Magnetization switching through giant spin–orbittorque in a magnetically doped topological insulatorheterostructure. Nature materials. 2014;13(7):699–704.

[131] Fan Y, Kou X, Upadhyaya P, Shao Q, Pan L, LangM, et al. Electric-field control of spin–orbit torquein a magnetically doped topological insulator. Naturenanotechnology. 2016;p. 352359.

[132] Spaldin NA, Fiebig M. The renaissance of magnetoelectricmultiferroics. Science. 2005;309(5733):391–392.

[133] Katsura H, Nagaosa N, Balatsky AV. Spin currentand magnetoelectric effect in noncollinear magnets.Physical review letters. 2005;95(5):057205.

[134] Kubacka T, Johnson JA, Hoffmann MC, Vicario C,De Jong S, Beaud P, et al. Large-amplitude spindynamics driven by a THz pulse in resonance with anelectromagnon. Science. 2014;343(6177):1333–1336.

[135] Battiato M, Carva K, Oppeneer PM. Superdiffusive spintransport as a mechanism of ultrafast demagnetization.Physical review letters. 2010;105(2):027203.

[136] Graves CE, Reid AH, Wang T, Wu B, de Jong S,Vahaplar K, et al. Nanoscale spin reversal by non-localangular momentum transfer following ultrafast laserexcitation in ferrimagnetic GdFeCo. Nature materials.2013;12(4):293–298.

[137] Rudolf D, Chan LOO, Battiato M, Adam R, Shaw JM,Turgut E, et al. Ultrafast magnetization enhancementin metallic multilayers driven by superdiffusive spincurrent. Nature communications. 2012;3:1037.

[138] Choi GMM, Min BCC, Lee KJJ, Cahill DG. Spin currentgenerated by thermally driven ultrafast demagnetiza-tion. Nature communications. 2014;5.

[139] Malinowski G, Dalla Longa F, Rietjens JHH, PaluskarPV, Huijink R, Swagten HJM, et al. Control of speedand efficiency of ultrafast demagnetization by directtransfer of spin angular momentum. Nature Physics.2008;4(11):855–858.

[140] Tows W, Pastor GM. Many-body theory of ultrafastdemagnetization and angular momentum transfer inferromagnetic transition metals. Physical reviewletters. 2015;115(21):217204.

[141] Eisebitt S, Luning J, Schlotter WF, Lorgen M, HellwigO, Eberhardt W, et al. Lensless imaging of magneticnanostructures by X-ray spectro-holography. Nature.2004;432(7019):885–888.

[142] Wang T, Zhu D, Wu B, Graves C, Schaffert S,Rander T, et al. Femtosecond single-shot imaging ofnanoscale ferromagnetic order in Co/Pd multilayersusing resonant x-ray holography. Physical reviewletters. 2012;108(26):267403.


[143] von Korff Schmising C, Pfau B, Schneider M, GuntherCM, Giovannella M, Perron J, et al. Imagingultrafast demagnetization dynamics after a spatiallylocalized optical excitation. Physical review letters.2014;112(21):217203.

[144] Buttner F, Moutafis C, Schneider M, Kruger B,Gunther CM, Geilhufe J, et al. Dynamics andinertia of skyrmionic spin structures. Nature Physics.2015;11(3):225–228.

[145] Eriksson M, Lindgren LJJ, Sjostrom M, Wallen E, RivkinL, Streun A. Some small-emittance light-source latticeswith multi-bend achromats. Nuclear Instruments andMethods in Physics Research Section A: Accelerators,Spectrometers, Detectors and Associated Equipment.2008;587(2):221–226.

[146] Bartosik MR, Bocchetta CJ, Borowiec P, Goryl P,Nietubyc R, Stankiewicz MJ, et al. SolarisNationalsynchrotron radiation centre, project progress, May2012. Radiation Physics and Chemistry. 2013;93:4–8.

[147] Reich ES. Ultimate upgrade for US synchrotron. Nature.2013;501(7466):148–149.

[148] Pellegrini C. A 4 to 0.1-nm FEL based on the SLAC Linac.In: Conf. Proc.. vol. 920224; 1992. p. 364–375.

[149] Hogan M, Pellegrini C, Rosenzweig J, Travish G,Varfolomeev A, Anderson S, et al. Measurements ofhigh gain and intensity fluctuations in a self-amplified,spontaneous-emission free-electron laser. Physicalreview letters. 1998;80(2):289.

[150] Ayvazyan V, Baboi N, Bahr J, Balandin V, Beutner B,Brandt A, et al. First operation of a free-electron lasergenerating GW power radiation at 32 nm wavelength.The European Physical Journal D-Atomic, Molecular,Optical and Plasma Physics. 2006;37(2):297–303.

[151] Altarelli M. The European X-ray free-electron laserfacility in Hamburg. Nuclear Instruments and Methodsin Physics Research Section B: Beam Interactions withMaterials and Atoms. 2011;269(24):2845–2849.

[152] Patterson BD, Abela R, Braun HH, Flechsig U, Ganter R,Kim Y, et al. Coherent science at the SwissFEL x-raylaser. New Journal of Physics. 2010;12(3):035012.

[153] Kang HSS, Kim KWW, Ko IS. Status of PAL-XFEL. In:SPIE Optics+ Optoelectronics. International Societyfor Optics and Photonics; 2015. p. 95120P–95120P.

[154] Allaria E, Callegari C, Cocco D, Fawley WM, KiskinovaM, Masciovecchio C, et al. The FERMIElettra free-electron-laser source for coherent x-ray physics: photonproperties, beam transport system and applications.New Journal of Physics. 2010;12(7):075002.

[155] Higley DJ, Hirsch K, Dakovski GL, Jal E, Yuan E, Liu T,et al. Femtosecond X-ray magnetic circular dichroismabsorption spectroscopy at an X-ray free electron laser.Review of scientific instruments. 2016;87(3):033110.

[156] Lutman AA, MacArthur JP, Ilchen M, Lindahl AO, BuckJ, Coffee RN, et al. Polarization control in an X-rayfree-electron laser. Nature Photonics. 2016;p. 468472.

[157] Galayda JN. The LCLS-II project. Proceedings of IPAC2014. 2014;.

arxiv:1611.07691v2 [cond-mat.mes-hall] 19 dec 2016x-ray imaging of spin currents and magnetisation...

Documents