lorentz microscopy

8/7/2019 LOrentz microscopy

1/8

4D Lorentz Electron Microscopy Imaging:Magnetic Domain Wall Nucleation, Reversal,and Wave VelocityHyun Soon Park, J. Spencer Baskin, and Ahmed H. Zewail*

Physical Biology Center for Ultrafast Science and Technology, Arthur Amos Noyes Laboratory of Chemical PhysicsCalifornia Institute of Technology, Pasadena, California 91125

ABSTRACT Magnetization reversal is an important topic of research in the elds of both basic and applied ferromagnetism. For thestudy of magnetization reversal dynamics and magnetic domain wall (DW) motion in ferromagnetic thin lms, imaging techniquesare indispensable. Here, we report 4D imaging of DWs by the out-of-focus Fresnel method in Lorentz ultrafast electron microscopy(UEM),with in situ spatial andtemporal resolutions.The temporal changein magnetization, as revealed by changes in image contrast,is clocked using an impulsive optical eld to produce structural deformation of the specimen, thus modulating magnetic eldcomponents in the specimen plane. Directly visualizedare DW nucleationand subsequent annihilation and oscillatory reappearance

(periods of 32 and 45 ns) in nickel lms on two different substrates. For the case of Ni lms on a Ti/Si3N4 substrate, under conditionsof minimum residual external magneticeld, theoscillation is associated with a uniquetraveling wave train of periodic magnetizationreversal. The velocity of DW propagation in this wave train is measured to be 172 m/s with a wavelength of 7.8m. The success of this study demonstrates the promise of Lorentz UEM for real-space imaging of spin switching, ferromagnetic resonance, and laser-induced demagnetization in ferromagnetic nanostructures.

KEYWORDS 4D imaging, magnetic domain dynamics, ultrafast electron microscopy, ferromagnetic materials

Ferromagnetism is an ancient but still active eld of study, covering phenomena such as the Zeemaneffect, giant magnetoresistance, and magnetic reso-nance imaging.1 - 3 Depending on the coerciveeld strength,three categoriesof ferromagnetic materials may be dened(soft, semihard, and hard) with a wide range of applicationsincluding data recording and storage.4 Magnetization rever-sal plays an essential role in many of these applications, andthis process and its dynamics have been the subject of extensive experimental and theoretical investigations, utiliz-ing a number of microscopy techniques that have beendeveloped for observing the magnetization state of ferro-magnetic materials.4 - 22

For studiesof the dynamics of magnetization reversalanddemagnetization, several techniqueswith the requisite timeresolution of femto- to nanoseconds have been imple-mented, including time-resolved magneto-optical Kerr mi-

croscopy, several variants of photoemission electron mi-croscopy, and scanning transmission X-raymicroscopy.10 - 15These rely on pulsed laser and synchrotron X-ray radiationsources which determine the ultimate time resolution andin specic applications have reached spatial resolutions aslow as 30 nm. As an example of the results obtained usingsuch techniques, magnetic domain wall (DW) velocities inNi80Fe20 permalloy under current-driven or magnetic-elddriven conditions have been measured in a broad range

from 3 to 1500m/s, depending on the current or the appliedmagnetic eld.16 - 18

Another technique among the repertoire of so-calledmagnetic imaging methods is Lorentz microscopy, whichcan be implemented in a standard transmission electronmicroscope by the Fresnel (out-of-focus) or Foucault (in-focus) method. This technique has been widely utilized toimage magnetic microstructures and the magnetizationprocess in situ,19 - 22 and a spatial resolution of 5 nm hasbeen reported for the Foucault method.21 In the Fresnelmethod (see below), domain walls are imaged, and theseplacea lower limitof tens of nanometers on the requirementfor spatial resolution. The Fresnel method has been used tocapture DW nucleation and motion at millisecond timeresolution (by coupling with video camera recording),22 butthehigh temporalresolutiondomainhasnotpreviously beenexplored.

4D electron microscopy, which was developed (seerefs 23- 29) for the observation of structural dynamicswith high spatiotemporal resolution, has been applied innumerous studies of various specimens (e.g., gold, graphite,silicon, iron, and carbon nanotubes) and for a range of phenomena (including transient andoscillatory crystal latticedeformation, morphological evolution, and phase transfor-mations), but none of these previous studies made use of magnetic imaging. In this contribution, we report, for therst time, the application of the Fresnel variant of 4D Lorentzelectron microscopy (Figure 1) which enables the visualiza-tion of magnetization dynamics induced by an impulsive

*To whom correspondence should be addressed, [email protected] for review: 08/12/2010Published on Web: 08/25/2010

pubs.acs.org/NanoLett

2010 American Chemical Society 3796 DOI: 10.1021/nl102861e | Nano Lett. 2010, 10 , 37963803


2/8

heating in soft ferromagnetic Ni thin lms. DW nucleation,oscillation, and wave propagation are directly imaged.

Microscopy and Materials. The specimens studied hereconsisted of polycrystalline thin lm layers of Ni on Ti (30nm/30 nm) deposited sequentially by electron beam evapo-ration onto two different types of substrate: Formvar mem-branes (30- 60 nm thick) on 200-mesh copper TEM grids;and a Si3N4 membrane (15 nm thick, 0.25 0.25 mm2

window). After the deposition of the twometal layers, squareislands ( 3.7 3.7 m2), as seeninFigure2,werepatternedby milling away the surrounding metal layers to form aframe of 3 m wide channels by the focused ion beam(FIB) method. The primary purpose of these islands was topin by the created defects the domain walls in distinct,reproducible patterns as the magnetization state evolvesthroughout each laser heating cycle; this is not the case incontinuous, homogeneous lms, in which domains nucleateat random. Note that for the Formvar substrate the FIBmilling completely removed the Formvar lm surroundingthe islands, leaving them sustained and connected with thesurrounding lm only by nanostructures like bridges and

webs (see Figure 2), while in the Si3N4 specimen the hardersubstrate was not fully milled away by FIB.

All experiments were performed with the second genera-tion of UEM(UEM-2) in thenanosecond stroboscopicmode,the operation of which has been described elsewhere.24,29Briey, the output (1064 nm) of a Q-switched Nd:YAG laserwas frequency-doubled to generate the 532 nm pump pulse,which was used to heat the specimen. The third harmonic

output of a second Q-switched Nd:YAG laser at 355 nm wasused to generate the electron probe pulse, which wasaccelerated to 200 keV. Here, all images were constructedstroboscopically at 1 kHz excitation frequency, providing a1 ms recovery time for the specimen between pulses. Therelative timing of pump and probe at the specimen wasestablished by triggering the two lasers by two indepen-dently programmable output pulses from a digital delaygenerator. For these experiments, the heating-pulse energyat the specimen was approximately 0.18J (Ni/Ti on Form-var) and 0.24 J (Ni/Ti on Si3N4) corresponding to peakuences of 6.4 and 8.5 mJ/cm2, respectively, for an esti-mated 50 m full-width-at-half-maximum Gaussian spot

FIGURE 1. 4D Lorentz electron microscopy. Schematic illustration of the deection of pulsed electrons on passing through a ferromagnetic

lm with the (saturation) magnetization M and thickness d . The magnetization is in the plane of the lm and reverses by 180 in alternatedomains. Domain walls appear as thick bright and dark lines in the out-of-focus image corresponding to this magnetization pattern, as shownin the frame labeled t ) 0. Magnetization ripple 35 is also illustrated. The frame at t ) t + is a schematic of the type of temporal change that can be observed in Ni. (See also Figure 5.)

2010 American Chemical Society 3797 DOI: 10.1021/nl102861e | Nano Lett. 2010, 10 , 3796-3803
http://pubs.acs.org/action/showImage?doi=10.1021/nl102861e&iName=master.img-000.jpg&w=299&h=347


3/8

size; the temporal width of the pulses was 10ns. Note that,under this focusing condition, the irradiance varies substan-tially across the observed specimen areas. The correspond-ing maximum local increase in sample temperature after thelaser pulse was estimated to be about 110 K (on Si3N4) atthe point of maximum irradiance, assuming equilibrationacross the 75 nm thickness of the entire heterogeneous lm.This estimate makes use of the heat capacities and opticalproperties of the materials.

In high magnication mode in our microscope, the speci-men is subjected to a magnetic eld generated by theobjective lens of up to 20 kOe, causing severe perturbationof the magnetic domain structure of magnetic materials. Toreduce the magnetic eld at the specimen position, experi-ments were conducted in low magnication mode (1000 or1500 , objective lens settings of 0% and 6%, respectively)with the main objective lens turnedoff. Foran objective lenssetting of 0%, the residual magnetic eld can be reduced toless than 2 Oe at the specimen.30 Under thiscondition, time-resolved magnetic imaging with femtosecond resolutionshould also be possible.

Time-Resolved Fresnel Images. Figure 1 illustrates thebasic concept of theFresnel methodof Lorentz microscopy,in which the locations of DWs are revealed in out-of-focusimages. The Lorentz techniques rely on deection of theaccelerated electrons by the magnetic Lorentz force:F b qvb Bb , where q is the electric charge,vb is the velocity, andBb isthe magnetic eld strength. The magnitude of the Lorentz

deection angle of electrons passing through a ferromag-netic thin lmof uniform magnetization andno applied eldis given by

where e is the electron charge, 4 M s is the saturationmagnetization, h is Plancks constant, is the electronwavelength, andd is the sample thickness.31 For example,the Lorentz deection angle for a 30 nm Ni lm (4 M s )6000 G) in our microscope (200 keV, ) 0.025 ) is 0.01mrad. Because the direction of the Lorentz force dependson that of the magnetization, the direction of electron beamdeection changes fromdomain to domain,creatingregionsbeyond the specimen of either deciency or excess of electrons (electron trajectories diverging or converging,respectively) at the position of a DW. Therefore, under adefocused condition, the so-called Fresnelmode, white andblack lines appear in the image depending on the sense of rotation of the magnetization andthesign of thedefocusing,i.e., whether over-focused or under-focused. (Characteristicpatterns of magnetization ripple also appear, as depicted intheschematic illustration; see below.)In contrast, in Foucaultmode, or in-focus, Lorentz imaging, an aperture is positionedto favor a particular direction of beam deection, thusenhancing the image contrast of areas of correspondingmagnetization, particularly for stronger magnetic materials.

An example of the effect of focus in Lorentz images,recorded 110 ns after a laser heating pulse, is shown in thetop row of Figure 2 for Ni/Ti on Formvar. Reversal of imagecontrast upon change from the over-focused to under-focused condition is apparent. The width of the white orblack lines depends on the width of the DW and the

magnitudeof thedefocusing. In the in-focus image of Figure2, no effect of the magnetization of the Ni lm is seen onimage contrast. Indeed, no contrast effects of any type are

FIGURE 2. 4D Lorentz images of Ni/Ti lms (30 nm/30 nm) on Formvar. Top row: Images captured at 110 ns delay, for in-focus, over-focus,and under-focus conditions, respectively. Note that the domain wall contrast is reversed between the over-focused and under-focused images.Prominent examples are circled. In the under-focused and over-focused images, weak ghost images of the bright pinning structure are alsoobserved. Bottom row: Time-dependent change of magnetic domain walls. In the image taken at - 9 ns, typical characteristic features of ripple are observed as white and black lines. At positive times, after the arrival of the clocking pulse (uence ) 6.4 mJ/cm 2) at t ) 0, dynamicsare observed in domain wall contrast (thicker bright or dark lines).

) e(4 M s)d /h (1)



4/8

observed or expected in in-focus images of the uniformpolycrystalline lm; thereforeallvariations in contrast in theout-of-focus imagesare interpreted as domainwall(or ripple)contrast.

In time-resolved imaging, a change in magnetization cancause nucleation, annihilation, or displacement of DWs, asdepicted in Figure 1 att ) t + . Three images representingthe observation of such changes following laser heating are

displayed in the bottom row of Figure 2. These images of Ni/Ti on Formvar were recorded with an objective lenssetting of 6% at different time delays relative to the heatingpulse. In the image taken at a delay of - 9 ns, white and blacklines appear as characteristic features which are typical of magnetic ripple as illustrated in Figure 1. At positive times,after the arrival of the clocking pulse att ) 0, nucleation andannihilation of domain walls are evident via changes inimage contrast; that is, zigzag DWs are observed to appearstrongly at 18 ns and to have largely disappeared by 36 ns.(This zigzag form of wall is known to be metastable sincelower energy congurations can be achieved by rearrange-

ment of the domain pattern.7

) Note that intensity changesdirectly represent changes in magnetization direction as afunction of time.

Contrast proles from the selected-area outlined by thewhite rectangles in the images of Figure 2 are given inFigure3 fora seriesof time delays, as indicated.Prolesmade fromimages recorded in two consecutive time scans are plottedtogether, demonstrating that the dynamics are reproducible.The proles reveal clearly the gradual appearance anddisappearance of the DWs. The time frame images fromFigure 2 were combined with others not shown to producea movie of the magnetization dynamics extending to longerdelay (Movie S1). In this movie, domain walls are seen to

appear and disappear periodically in the same location,indicating that an oscillatorybehavior is initiated by the laserheating. Further details of this behavior are given below inconnection with the discussion of Figure 5.

A similar DW oscillation following laser excitation wasstudied in a different area on the same substrate, also withan objective lens setting of 6%. The time dependence of DWs from a Lorentz UEM time scan is represented by theplot of selected-area integrated image intensity shown in

Figure 4. The selected area for the plotted intensity isoutlined by a white rectangle in oneout-of-focus image fromthe scan shown at top left in the gure. The intensityoscillates with a period of 32 ns. Interestingly, under thesameexcitation condition, several small structural remnantsof the FIB milling process surrounding the adjacent islandwere observed to undergo mechanical motion. In-focusimages at higher magnication (3600 , objective lens 86%),including that shown at the top right of Figure 4, wererecorded in time scans, and a movie was made from thetime frames (Movie S2). Analysis by selected-area imagecross correlations, such as the one plotted in Figure 4 (seecaption), revealed that structures differing widely in size and

FIGURE 3. Time-dependent change of magnetic domain walls. Fromproles corresponding to the selected area outlined by whiterectangles in the images of Figure2, the nucleation and annihilationof domain walls can be visualized. Proles from two consecutivescans are shown.

FIGURE 4. Domain wall oscillation and mechanical motion of Ni/Tilms on Formvar. The top left image is one frame from an out-of-focus time scan (magnication 1500 , objective lens 6%). Time-dependent change of domain walls during the full scan can be seenby the selected-area image intensityof thearea indicated by a whiterectangle in the out-of-focus image. At top right is shown one imagefrom an in-focus time scan (magnication 3600 ), and the timedependence of the image cross correlation (reference at t ) - 50ns) of a selected-area (black rectangle) indicated in that image is

plotted.Excitation uence for bothscans was 6.4 mJ/cm2

. Oscillationperiods p are measured to be 32 ns in both time series.

http://pubs.acs.org/action/showImage?doi=10.1021/nl102861e&iName=master.img-003.jpg&w=191&h=273http://pubs.acs.org/action/showImage?doi=10.1021/nl102861e&iName=master.img-002.jpg&w=194&h=201


5/8

form exhibited oscillations at frequencies near 30 MHz. Asis clearly seen, the phase and frequency of the plottedmechanical and DW oscillations match closely.

Mechanism of Nucleation and Reversal. Given theseobservations, we conclude that a single mechanical reso-nance29 is initiated in the lm by the laser pulse and thatthis resonance drives forced oscillations in both the DW

dynamics andthe mechanical motion of thenanostructures.That the resonance is mechanical in nature rather thanhaving a magnetic origin is indicated by the constantfrequency under the very different magnetic conditions atthe specimen for 6% and 86% settings of the objective lens.On the basis of the elastic and mechanical properties of Niand Ti,32 a fundamental resonance period of tens of nano-seconds could only be consistent with the exural reso-nance33,34 of a 60 nm thick plate when the lateral dimen-sions are of the order of those of the island (3.7m), so wepropose that such a vibrationof the islandstructureprovidesthe driving force for the observed dynamics. It is initiated

when the nanosecond laser heating pulse induces an impul-sive lm distortion or bulging, following conversion of theabsorbed photon energy to thermal energy on the picosec-ond time scale, as observed previously in graphite and goldlms.24,25,29 The bilaminar nature of the Ni/Ti lm willcontribute to such a exing motion upon heating, even intheabsenceofothermechanicalconstraintsonlmexpansion.

The relevant changes in magnetization associated withthe dynamics of Figures 2 and 3 are summarized in Figure5, in which expanded views of two characteristic timeframes are shown. As noted above, a magnetic ripplestructure appears in the image of the room temperaturespecimen at negative time.35 Knowing that the ripple struc-

ture runs perpendicular to the average magnetization direc-tion, the local magnetization direction can be derived fromit, as illustrated in the drawing below the image. In thesecond image, the transient zigzag domain walls are shownat the time of theirmaximum intensity. Theseareassociatedwith a magnetization direction (red arrows in the lowerpanel) that is perpendicular to that of the equilibrium state.

To explain the above changes, we propose the followingmechanism for DW nucleation and oscillation. As shown inthe schematic illustration at the right of Figure 5, it isassumed that the resonant bending of the island induces anoscillatory physical distortion in the adjacent lmby meansof the connecting mechanical support structures. By suchdriven oscillation in the local surface tilt, part of the residualinstrumental magnetic eld perpendicular to the specimen(H 0) can be introduced into the specimen lm plane as avariable eldH | (t ). For a tilt angle of (t )

If the maximum value of H | is above the coercive eldof Ni,the formation of domain walls can be induced as shown inthe gure. Note that the magnetization direction is changedboth parallel and antiparallel to the direction of H | .

By tilting the specimen (without laser excitation) from itsusual orientation normal to the microscope axis, we testedthe effect of a projection of the residual axial magnetic eldinto the lm plane. Under the conditionof objective lens 0%,domain wall changes could not be induced in our specimeneven if the sample was tilted 10. Thus, for the axial eld at

FIGURE 5. Domain wall nucleation and annihilation. Left: Expanded views of two images taken from Figure 2 are shown. The correspondingmagnetization is shown in the lower panels. Right: Schematic illustration showing a possible mechanism of domain wall nucleation. A laserheating pulse induces surface bulging and resonant vibration of an adjacent island structure (not shown), followed by mechanical vibrationof the thin lm. A magnetic eld component H | (t ) is introduced into the tilted specimen lm plane, with magnitude given by H 0 sin (t ),where H 0 is a residual eld on the microscope axis. Note that the magnetization distribution of the ripples changes parallel and antiparallelto the direction of the applied magnetic eld, resulting in the nucleation of zigzag domain walls below the maximum tilt angle max .

H | (t ) ) H 0 sin (t ) (2)



6/8

this condition (specied at< 2 Oe), a 10 tilt gives an in-planemagnetic eld below the coercive eld of Ni. On the other

hand, by sample tilting of only 2, domain walls wereseverely changed in the case of an objective lens setting of 6%. Thus, for the 6% setting, if localized tilting followingimpulsive heating is as large as 2, the in-plane magneticeld induced will be larger than the coercive eld of Ni andmay trigger the periodic magnetization reversal observed.It should be noted that such reversal dynamics were notobserved in our experiments using an objective lens settingof 0%.

Alternatively, in the absence of strong tilting of the lmat the location of the DW nucleation, a sufciently strongoscillatingH | (t ) at that location may result indirectly from

the elds associated with the spatially oscillating magnetiza-tion directionof the tilting island itself, even if the islanddoesnot undergo magnetization reversals in response toH 0. Theamplitude of deformation is expected to be much greaterin the resonant motion of the island structure than in anydriven nonresonant response of the main lm. Given thatthe DW shape is generally determined by the balancebetween the Zeeman energy (external applied magneticeld) and the DW energy, it seems reasonable to expect thatmetastable zigzag walls are indeed a characteristic featureof the application of a transient in-plane magnetic eld.

Domain Wall Wave Propagation. Another series of 4DLorentzelectronmicroscopyexperimentswas performed on

Ni/Ti onSi3N4 substrate. In these, DW wave propagation wasobserved, and measurements are displayed in Figure 6

(Movie S3). The wavepropagates downward within theareaoutlined by the long rectangular box shown in the upper leftimage in the gure. Selected-area-image dynamics as rep-resented by integrated contrast intensities from the twosmall shaded squares shows strong oscillatory behaviorwithposition-dependent phase, indicative of the passage of thewave (Figure 6, center). These oscillations continued to atleast 1 s, with a time constant of damping of amplitudeestimated at 2- 3 s. The dynamics also includes a nonoscil-latory background offset aftert ) 0, which decays back tothenegative time baselinevalue with a time constant of 16s, consistent with the time scale for diffusive cooling

estimated for such a lm. In the difference image at bottomleft of the gure, white, black and gray contrast indicatesan increase, decrease, or no change in imagecontrast at thespecied time relative to the reference image at negativetime. Thus, contrast variations in the lm (but not in the FIB-milled areas; see below), directly correspond to changes inmagnetization direction.

The time evolution of the contrast prole within the largewhite rectangle in the upper image of Figure 6 is shown atthe right of the gure, and clear diagonal lines of subtlecontrast variation represent thewave motion. Theperiod pand DW velocity are measured to be 45 ns and 172 m/s,respectively. The wave appears to be initiated by a mechan-

FIGURE 6. Images, time dependence of selected-area intensity, and domain wall (DW) propagation forNi/Ti thin lm on Si 3N4 substrate (uence) 8.5 mJ/cm 2). In the images, white and black lines correspond to DWs. The difference image at lower left shows only features that havechanged between the reference time ( - 50 ns) and 205 ns, including a dramatic displacement toward the upper left of the top FIB-milledisland. The intensity plots in the middle panel are the integrated contrast intensities of selected areas (25 23 pixels, 7.7 pixels ) 1 m)indicated in the top image by boxes of corresponding shading. Both plots are intensity changes relative to - 50 ns, with the blue plot offset for clarity. The DW oscillation period is 45 ns (22 MHz). Note the out-of-phase nature of the oscillations for the traveling wave and the intensitystep at t 10 ns. The plot at right displays the time evolution of the intensity prole along the long axis of the rectangular box in the topimage. Propagation of a DW wave train is represented by the series of parallel diagonal contrast lines (directions indicated by the black arrows)from which a DW velocity of 172 m/s is measured. The wavelength is measured as 7.8 m. The horizontal red and blue arrows indicate thespatial positions at which the time proles shown in the middle panel were taken.



7/8

ical shock att ) 0 that also causes an impulsive shift in therelative positions of adjacent FIB holes (see the area indi-cated by the yellow dotted circle in the difference image of Figure 6). This shock is presumed to set up a strong me-chanical resonance in the island much as occurred in theFormvar specimen, though with a modied frequency dueto thedifferent substrateanddifferingboundary conditions.

In theoretical models of external-eld-driven magnetiza-tion reversal propagation,7,36 an expression forDW velocityin the low eld limit is

Here m is the DW mobility, is the electron gyromagneticratio, is the domain wall width,R is the Gilbert dampingparameter, H is the in-plane applied eld, andH c is the

coercive eld for DW motion. This expression is valid forelds up to the critical driving eld of Walker, with atheoretical value of H ) 2 R Ms. Upon substitution oneobtains

For Ni ( 72 nm) and Ni80Fe20 ( 20 nm) lms,max iscalculated to be about 3800 m/sand1760 m/s, respectively.The experimental measurements of the DW range from 3to 1500 m/s in Ni80Fe20 depending on the experimental

technique and applied eld, and the critical eld has beenfound to be much lower than predicted,16 - 18 resulting in alow eldmax of 75 m/s in Ni80Fe20.

In the experimental technique reported here, under aresidual total magnetic eld of less than 2 Oe (objective lens0%) and much smaller in-plane applied eld, we directlymeasured of 172 m/s, a frequencyf of 22 MHz, and awavelength of 7.8mfortheSi3N4 substrate. This measured is several tens of times larger than the calculated valuefrom eq 3, even with a large induced tilt. Moreover, in ourcase the magnetization reversal occurs as a periodic wavetrain, with themagnetization repeatedly switching direction,

unlike the single reversal to a lower energy state in the caseof theexternal-eld-driven domain wall propagation. It thusappears that the wave, generatedhere bya local, periodicallyvarying perturbationcaused by mechanical oscillation of theisland structure, travels across thelmas a self-propagatingdisturbance in the magnetic structure. The wave velocity isnot considered to be determined by only the mechanicalelastic properties of the lm because the speed of sound inNi and Ti is known to range from 3000 to 6070 m/s forvarious wave types and geometries,37 which is a minimumof 17 times larger than our measured velocity.

The origin of the postulated self-propagation of the wavetrain observed here can now be addressed. In the static

equilibrium state of magnetization beforet ) 0, eachindividual grain has its own magnetic moment alignment,controlled by its particular crystal orientation and the col-lectively determined magnetic environment, resulting in acharacteristic ripple structure, as illustrated in Figures 1 and5. As the temperature increases immediately after laserexcitation, more and more magnetic moments deviaterandomly from the common direction, thereby increasingthe internal energy and reducing the net magnetization.Heating also causes a strong local perturbation of the equi-librium magnetic environment in the immediate vicinity of the island as a result of a resonant mechanical distortion of the type previously described. Thus individual magneticmoments will undergo a characteristic precessional motionundertheinuenceofthetorqueexertedonthemagnetization.

The precession rate determines the time scale for themagnetization reversal of each grain. Each reversed mag-netization inuences neighboring grains, causing the mag-netization response to propagate at a rate dependent on thestrength and range of the induced elds. If each magnetiza-tion reversal step occurs on a time scale of 200 ps, theestimated switching time for coherent magnetization rota-tion,38 then our measured velocity corresponds to a rangeof strongly coupled magnetization extending over about 3grain diameters of 10 nm each. This range of magnetizationcorrelation can be related to the domain wall width, and itsrole in the wave velocity here parallels that of in thevelocity of eq 3, though apparently with a local, self-gener-ated, effective driving eld.

Conclusion. We have shown here that, with 4D Lorentzelectron microscopy, magnetization reversal dynamics anddomain wall (DW) velocity in soft ferromagnetic thin lmscan be successfully studied with spatiotemporal resolution.Periodic perturbations created by surface deformation fol-lowing the impulsive heating causes introduction of a mag-netic eld component into the lm plane, which triggers thenucleation of zigzag domain walls frommagnetization ripplestructures, followed by subsequent oscillation. DW wavetrain propagation was also observed under minimum exter-naleld conditions,with velocity andwavelength measuredto be 172 m/s and 7.8m, respectively. Further studies of

spin switching, precessional motion, and DW properties invarious ferromagnetic materials are expected with LorentzUEMin the future.Possible applicationscould include studiesto improve the speed of magnetic recording devices.

Acknowledgment. This work was supported by the Na-tional Science Foundation and Air Force Ofce of ScienticResearch in theGordonandBetty Moore Center forPhysicalBiology at Caltech.

Supporting Information Available. Videos showing timedependent change of magnetic domain walls, mechanicalmotion of Ni/Ti nanostructures, and domain wall propaga-

) m(H - H c) ) /(H - H c) (3)

max (4 M s )/2 (4)



8/8

tion for Ni/Ti thin lm on Si3N4 substrate. This material isavailable free of charge via the Internet at http://pubs.acs.org.

REFERENCES AND NOTES(1) Zeeman, P.Philos. Mag. 1897 , 44, 5566.(2) Fert, A.; Grunberg, P.Les Prix Nobel. The Nobel prizes 2007 ;

Grandin, K., Ed.; Nobel Foundation: Stockholm, 2008.(3) Rugar, D.; Yannoni, C. S.; Sidles, J. A.Nature 1992 , 360 , 563

566.(4) Chikazumi, S.Physics of ferromagnetism ; Oxford Science Publica-

tions: Oxford, 1997.(5) McCurrie, R.A.Ferromagnetic Materials: structure and properties ;

Academic Press: London, 1994.(6) Kronmuller, H.Micromagnetism and the microstructure of ferro-

magnetic solids ; Cambridge University Press: Cambridge, 2003.(7) Hubert, A.; Schafer, R.Magnetic domains: the analysis of magnetic

microstructures ; Springer-Verlag: Berlin, 1998.(8) Freeman, M. R.; Choi, B. C.Science 2001 , 294 , 14841488.(9) Magnetic imaging and its applications to materials ; Graef, M. D.,

Zhu, Y., Eds.; Academic Press: London, 2001.(10) Acremann, Y.; Back, C. H.; Buess, M.; Portmann, O.; Vaterlaus,

A.; Pescia, D.; Melchior, H.Science 2000 , 290 , 492495.(11) Choi, B. C.;Belov, M.; Hiebert, W. K.; Ballentine, G. E.;Freeman,

M. R.Phys. Rev. Lett. 2001 , 86 , 728731.(12) Choe, S.-B.; Acremann, Y.; Scholl, A.;Bauer,A.;Doran,A.;Stohr,

J.; Padmore, H. A.Science 2004 , 304 , 420422.(13) Van Waeyenberge, B.; et al.Nature 2006 , 444 , 461464.(14) Chembrolu, V.; et al.Phys. Rev. B 2009 , 80 , No. 024417.(15) Heitkamp, B.; Kronast, F.; Heyne, L.; Durr, H. A.; Eberhardt, W.;

Landis, S.; Rodmacq, B.J. Phys. D: Appl. Phys. 2008 , 41 , 164002.(16) Atkinson, D.; Allwood, D. A.; Xiong, G.; Cooke, M. D.; Faulkner,

C. C.; Cowburn, R. P.Nat. Mater. 2003 , 2, 8587.(17) Beach, G. S. D.; Nistor, C.; Knutson, C.; Tsoi, M.; Erskine, J. L.

Nat. Mater. 2005 , 4, 741744.(18) Yamaguchi, A.; Ono, T.; Nasu, S.; Miyake, K.; Mibu, K.; Shinjo,

T. Phys. Rev. Lett. 2004 , 92 , No. 077205.(19) Tonomura, A.Proc. Natl. Acad. Sci. U.S.A. 2005 , 102 , 14952

14959.

(20) McCartney, M. R.; Dunin-Borkowski, R. E.; Scheinfein, M. R.;Smith, D. J.; Gider, S.; Parkin, S. S. P.Science 1999 , 286 , 13371340.

(21) Majetich, S. A.; Jin, Y.Science 1999 , 284 , 470473.(22) Park, H. S.; Park,Y.-G.; Gao, Y.;Shindo,D.; Inoue, M.J. Appl. Phys.

2005 , 97 , No. 033908.(23) Zewail, A. H.; Thomas, J. M.4D electron microscopy: imaging in

space and time ; Imperial College Press: London, 2010.(24) Barwick, B.; Park, H. S.; Kwon, O.-H.; Baskin, J. S.; Zewail, A. H

Science 2008 , 322 , 12271231.(25) Kwon, O.-H.; Barwick, B.; Park, H. S.; Baskin, J. S.; Zewail, A. H

Nano Lett. 2008 , 8, 35573562.(26) Yurtsever, A.; Zewail, A. H.Science 2009 , 326 , 708712.(27) Park, H. S.; Kwon, O.-H.; Baskin, J. S.; Barwick, B.; Zewail, A. H

Nano Lett. 2009 , 9, 39543962.(28) Barwick, B.;Flannigan, D. J.;Zewail, A. H.Nature 2009 , 462 , 902

906.(29) Park, H. S.; Baskin, J. S.; Barwick, B.; Kwon, O.-H.; Zewail, A. H

Ultramicroscopy 2009 , 110 , 719.(30) Shindo, D.;Oikawa,T.Analytical electronmicroscopy for materials

science ; Springer-Verlag: Tokyo, 2002.(31) Chapman, J. N.J. Phys. D: Appl. Phys. 1984 , 17 , 623647.(32) Ledbetter, H. M.; Reed, R. P.J. Phys. Chem. Ref. Data 1973 , 2,

531618.(33) Laura, P. A.; Saffell, B. F., Jr.J. Acoust. Soc. Am. 1967 , 41 , 836

839.

(34) Rao, S. S.Mechanical Vibrations ; Addison-Wesley: Reading, MA,1986.(35) Ripple structure caused by variation in magnetization direction

reects the randomly oriented polycrystalline nature (averagegrain size of 10 nm) of Ni lm and the crystal anisotropy of individual grains. The mean wavelengthL of longitudinalmagnetization ripple is measured in the current specimens tobe 900 nm, a value of the same order of magnitude asdetermined by experiment and theory for soft ferromagneticthin lms.39 In the thin lm on 200-mesh copper grid, themagnetization direction tends to be oriented parallel to theedges of the defect structures created by FIB in order tominimize the stray eld energy.

(36) Schryer, N. L.; Walker, L. R.J. Appl. Phys. 1974 , 45 , 54065421.(37) CRC Handbook of Chemistry and Physics , 82nd ed.; Lide, D. R.,

Ed.; CRC Press: Boca Raton, FL, 2001; p14 -41.

(38) Gerrits, T.; van den Berg, H. A. M.; Hohlfeld, J.; Bar, L.; Rasing, T.Nature 2002 , 418 , 509512.(39) Hoffmann, H.J. Appl. Phys. 1964 , 35 , 17901798.


lorentz microscopy

Documents