Hawkes P.W., Spence J.C.H. (Eds.) Science of Microscopy. V.1 and 2

Подождите немного. Документ загружается.

Chapter 9 Photoemission Electron Microscopy (PEEM) 675

used as a diagnostic to test and optimize the objective lens and the ﬁ rst

half of the beam separator. In this operation mode, the mirror acts as a

unipotential lens and the electron beam passes though a 500-µm hole

in the reﬂ ecting electrode.

The PEEM-3 electron mirror has four rotationally symmetric elec-

trodes, and the reﬂ ecting electrode is a segment of a sphere with a radius

of 5.6 mm. The inner electrode is held at ground voltage, while the poten-

tials of the other electrodes provide three free knobs to set the focal

length, the chromatic aberration correction, and the spherical aberration

correction of the mirror. Figure 9–16 shows the equipotential distribu-

tion and correction region of the mirror. This region is sufﬁ ciently large

to correct the microscope aberrations in various modes of operation.

4.3 The Beam Separator

The task of the magnetic beam separator is to separate the aberration-

corrected beam leaving the mirror from the beam that enters the mirror.

A separator with very small aberrations is critical for the performance

of an aberration-corrected PEEM microscope. Two different separator

designs have been considered and are currently being implement. The

SMART microscope uses a triple-bend achromat optic in each of the

four sectors of the separator magnet, which has a side length of 280 mm.

The route to this type of separator was ﬁ rst identiﬁ ed by Rose and

Preikszas

and further developed in the work of Müller et al. (1999)

and Wu et al. (2004). The design uses a highly symmetric beam path

to cancel a maximum number of aberrations. It is very compact but

Figure 9–15. Layout of the aberration-corrected PEEM-3 at the Advanced Light Source. An electron

mirror is used as an aberration corrector.

676 J. Feng and A. Scholl

Figure 9–17. Schematic layout and principal rays in the x–z (top) and y–z (bottom) plane of one 90

sector of the PEEM-3 separator. Two 90

bends transport the beam into the mirror and back to the

projector optics. One dipole (DI), two round lenses (RL), and four quadrupoles (EQ) are used.

Figure 9–16. Aberration correction by a tetrode electron mirror. Left panel, equipotential contours.

Right panel, aberration region that can be corrected by the mirror.

poses high demands on tolerances, and on mechanical and electrical

stability (Hartel et al., 2002).

Figure 9–17 shows the 90° beam separator that will be used in the

Advanced Light Source PEEM-3. The separator consists of three round

Chapter 9 Photoemission Electron Microscopy (PEEM) 677

lenses, one dipole magnet, and six electrostatic quadrupoles. Round

lenses are chosen to be the main focusing optics. Weak electrostatic

quadrupoles are used to form a stigmatic image after each 90° bend.

Each 90° bend is mirror symmetric about the center and acts as a tele-

scope in the x–z and the y–z plane. The linear dispersion and the

second-order axial aberrations, which are not corrected within each 90°

bend, cancel to zero after the second pass through the separator,

because of the negative unity magniﬁ cation of the mirror. Calculations

show that the chromatic and spherical aberrations of PEEM-3 and the

SMART separator are of the same order (Wan et al., in preparation).

The PEEM-3 beam separator uses electrostatic lenses instead of fringe

ﬁ elds for image focusing and is robust regarding the details of the

magnetic ﬁ eld in the separator. A simple dipole magnet can be used,

and the demands on the power supply stability and mechanical toler-

ances are relatively low.

4.4 Resolution and Transmission of an

Aberration-Corrected PEEM

Aberration correction signiﬁ cantly improves both the spatial resolution

and the transmission. The resolution of PEEM-3 was modeled in a way

similar as that of PEEM-2. An ensemble of electrons with different

initial energy and emission angles was sent through a model of the

electron optics, intensity weighted with the probability of each emission

energy and angle. The diffraction effect was calculated for each electron

in the ensemble and the result was summed up to yield the point-spread

function. The resolution was deﬁ ned as the distance between the 15th

percentile and 85th percentile of the point-spread function. The resolu-

tion and transmission of PEEM-3 are shown in Figure 9–18 in compari-

son with the uncorrected PEEM-2 (Feng et al., 2004). PEEM-3 is designed

to have two main operation modes, a high-esolution mode and a high-

ﬂ ux mode. Operating at 20 kV and 2 mm working distance, the resolu-

Figure 9–18. Comparison of calculated resolution versus transmission of the

Advanced Light Source PEEM-2 and PEEM-3. The acceleration potential is

20 kV and the sample–lens distance is 2 mm.

678 J. Feng and A. Scholl

tion for 100% transmission (no aperture used) reaches 50 nm using the

mirror corrector, a signiﬁ cant improvement over PEEM-2, which reaches

only 440 nm. Using a small aperture the optimum resolution will be

5 nm at 2% transmission, as opposed to 20 nm at 1% transmission for

PEEM-2. For any chosen spatial resolution, a signiﬁ cantly larger aper-

ture can be used, resulting in a large increase in efﬁ ciency over PEEM-2.

An aberration-corrected X-PEEM such PEEM-3 will have signiﬁ cantly

improved ultimate resolution and transmission, key attributes for the

imaging of nanostructures (Bauer, 2001a, 2001b).

5 Application: Magnetic Domain Imaging

Magnetic materials have applications in key areas of information tech-

nology, mainly because of low cost and nonvolatility, the ability to

retain information without a power source. Magnetic hard disks are

the dominant technology in data storage. New applications for mag-

netic materials emerge when magnetic and electronic properties are

concurrently utilized (Wolf et al., 2001), for example, in magnetic

random access memory (MRAM) chips. Magnetic storage technology

relies on structures with lateral dimensions of well below 1 µm

. The

storage density on commercial hard disks in 2004 was about 60 Gbit/

, corresponding to an area of 0.01 µm

per bit. At a typical bit aspect

ratio of about 10 : 1 this corresponds to a bit length of 30 nm and a track

width (or bit width) of 300 nm. In demonstrations and in some products

even smaller dimensions have been reached. Progress in device devel-

opment, and also in the fundamental understanding of the properties

of magnetic material, depends on our ability to characterize magnetic

materials on a nanoscale.

5.1 Magnetic Circular and Linear Dichroism

X-ray magnetic circular dichroism (XMCD) is a standard method to

study magnetic thin ﬁ lms and surfaces (Rortright et al., 1999). The wide

availability of polarized and tunable X-rays from synchrotron sources

has played an important role in the success of X-ray dichroism tech-

niques. In the past 15 years there has been great progress following the

ﬁ rst XMCD spectroscopy measurements at the important transition

metal L edges in 1987 by Schütz et al. (1987), the ﬁ rst imaging of ferro-

magnetic domains in 1993 by Stöhr et al. (1993), and the ﬁ rst imaging of

antiferromagnetic domains by Stöhr et al. (1999) and Scholl et al. (2000).

Circularly polarized X-rays measure the direction of the atomic mag-

netic moment of a ferromagnet relative to the polarization vector of the

X-rays. In the presence of spin orbit interaction the photon angular

momentum is transferred to the angular momentum of the photoex-

cited electron, in particular the electron spin, which senses the differ-

ent density of empty states in each spin channel. An imbalance of the

occupation of the majority and minority states is characteristic for a

ferromagnet. Strong XMCD effects (up to 40%) appear at the L edges

(2p → 3d transition) of the transition metal ferromagnets Fe, Co, and

Ni, since the d states carry most of the magnetic moment. The XMCD

spectrum is deﬁ ned as the difference spectrum of two absorption

Chapter 9 Photoemission Electron Microscopy (PEEM) 679

spectra acquired with either opposite polarization or opposite magne-

tization. The XMCD effect is opposite in sign at the L

and L

edge

because of the opposite sign of the spin-orbit coupling in the 2p states:

l + s for 2p

3/2

, and l − s for 2p

1/2

. The different coupling gives rise to a

unique feature of XMCD, its ability to separate spin and orbital moment.

The spin momentum is proportional to the difference of the integrated

XMCD intensity at the L

and the L

edge and the orbital momentum

is proportional to the sum. Sum rules have been developed that are

used to quantitatively determine the spin and orbital magnetic moment

per atom (Thole et al., 1992; Wu et al., 1992; Carra et al., 1993). The angle

and magnetization dependence of XMCD in the total absorption signal

is approximately given by I

XMCD

∼ cos α〈M〉

, with α denoting the angle

between the X-ray helicity vector Ε (parallel to the X-ray propagation

direction) and the magnetization Μ. Figure 9–19 shows X-ray absorp-

tion spectra of a Co ﬁ lm, which was magnetized parallel (solid line)

and antiparallel (dotted line) to the X-ray polarization. The spectra

were obtained by measuring the drain current on the sample using a

sensitive picoammeter. The drain current is also called total electron

yield (TEY) because it is equal to the total number of electrons that

leave the sample in response to the photoexcitation. In the lower part

of Figure 9–19, the XMCD difference spectrum is shown.

Linearly polarized X-rays probe the anisotropy of the electronic and

magnetic structure and are particularly useful for the study of antifer-

romagnets, which do not possess a macroscopic ferromagnetic moment.

In an antiferromagnet with a collinear magnetic structure, linearly

polarized X-rays distinguish between a linear polarization vector

Figure 9–19. Co X-ray magnetic circular dichroism (XMCD) spectra acquired

for a magnetization M parallel (dotted line) and antiparallel (solid line) to the

X-ray polarization σ. At the bottom the difference spectrum is shown.

680 J. Feng and A. Scholl

aligned with the magnetic spin axis and a linear polarization vector

perpendicular to the magnetic spin axis (Ruiper et al., 1993; Alders et

al., 1998; Lüning et al., 2003). In transition metal oxides, such as NiO,

CoO, Fe

, Fe

, LaFeO

, and others, an X-ray magnetic linear dichro-

ism (XMLD) of up to 10% has been observed. Linearly polarized X-rays

probe the angle α between the linear X-ray polarization vector Ε and

the orientation of the magnetization axis A

. The angle and magnetiza-

tion dependence of the XMLD intensity is approximately given by I

XMLD

∼ (1 − 3 cos

α)〈M

〉

. Here 〈M

〉

is the statistical average of the squared

moment at the temperature T.

5.2 Imaging of Ferromagnetic Domains

A subtle balance of energies in a magnetic material causes the formation

of magnetic domains. Contributions to the total energy are the exchange

energy, the magnetic anisotropy, and the dipolar energy. Dipolar ﬁ elds

are the driving force behind domain creation. Domains minimize the

magnetic stray ﬁ eld leaving the sample and therefore minimize the

dipolar energy. Exchange and anisotropy ﬁ elds resist domain formation

because they usually favor aligned spins. In magnetic thin ﬁ lms inter-

face effects such as interface exchange coupling, interlayer coupling, and

interface anisotropy alter this delicate balance and lead to new magnetic

phases. Transitions between different magnetic phases occur as a func-

tion of ﬁ lm thickness, interlayer thickness, magnetic ﬁ eld, and tempera-

ture, and are extensively studied because of the interesting physics

underlying the often complex phase diagrams. High-resolution mag-

netic imaging is an important tool to study magnetic domains and phase

transitions. X-ray techniques are especially powerful because they allow

us to separate the magnetic state of individual layers in a multilayer,

using element-speciﬁ c imaging. For a review of domain imaging using

X-PEEM see Schneider and Schönhense (2002).

As an example, the in- versus out-of-plane transition of a thin face-

centered cubic (fcc) Fe ﬁ lm on a ﬁ ve-monolayer Ni/Cu(001) substrate

is visualized in Figure 9–20 (Wu et al., 2004). The Ni layer produced

an in-plane exchange ﬁ eld that lowered the critical thickness at which

the Fe spin reorientation occurred to about 2.6 monolayers. The experi-

ments were conducted on an Fe wedge, which was grown under ultra-

high-vacuum (UHV) conditions. The shown images were acquired in

the transition region of in- and out-of-plane magnetization. XMCD

images were obtained by dividing a PEEM image taken at the Fe L

edge by an image taken at the L

edge. Since the magnetic dichroism

has an opposite sign at these edges, the procedure enhanced the mag-

netic contrast and suppressed nonmagnet effects. Below the transition

thickness, stripe domains formed, which had an out-of-plane magne-

tization. Approaching the transition, these domains shrank in size.

Large in-plane domains appeared above the transition thickness. The

experimentally observed thickness dependence of the stripe domain

width could be explained within a theoretical model that took into

account the exchange interaction, the anisotropy, the dipolar ﬁ eld, and

the virtual ﬁ eld created by the Ni layer (Wu et al., 2004).

Chapter 9 Photoemission Electron Microscopy (PEEM) 681

5.3 Imaging of Antiferromagnetic Domains

Antiferromagnetism is a material property that occurs in important

material classes: in transition metal oxides, for example, Ni oxide, Co

oxide, and Mn oxides, as well as in metallic alloys, for example, Mn

alloys. The unidirection al exchange coupling at the interface between

an antiferromagnet and a ferromagnet has found particular attention,

called exchange bias (Meiklejohn and Bean, 1956; Kiwi, 2001;

Berkowitz and Takano, 1999; Nogues and Schuller, 1999). Exchange

bias leads to a pinning of the magnetization of the ferromagnet after

annealing of the antiferromagnet/ferromagnet structure in a magnetic

ﬁ eld. It has found application in technology where it is used to pin the

magnetization of a ferromagnetic reference layer in devices that are

used as magnetic ﬁ eld sensors, e.g., in hard disk read heads. X-ray

spectroscopy and X-ray microscopy have been instrumental in study-

ing and understanding the microscopic magnetic structure of such

ferromagnet/antiferromagnet interfaces, because of the sensitivity of

X-ray spectroscopy and microscopy to antiferromagnetic order (Stöhr

et al., 1999; Scholl et al., 2000, 2001, 2004a, 2004b; Ohldag et al., 2001a,

2001b, 2003; Kuch et al., 2002; Ofﬁ et al., 2003a, 2003b, 2004).

LaFeO

is an antiferromagnet with a bulk Néel temperature of

740 K. Figure 9–21 shows local X-ray absorption spectra and images

acquired using X-PEEM of an epitaxially grown 40-nm LaFeO

ﬁ lm

on an SrTiO

(001) substrate (Scholl et al., 2000). The spectra were

acquired using linearly polarized X-rays. XMLD effect is visible at the

3/2

edges and appears as a variation in the L edge ﬁ ne structure.

Images acquired at energy positions labeled A and B in the spectrum

show a dark–bright contrast that due to antiferromagnetic domains

2.8 ML2.6 ML2.4 ML

µm

Domain width (µm)

Fe thickness (ML)

2.3 2.4 2.5 2.6 2.7

(c)(b)

(a)

Figure 9–20. (a) Fe stripe domain structure in Fe/Ni/Cu(001) at the spin

reorientation transition from an out-of-plane magnetization for small Fe thick-

ness to an in-plane magnetization for larger thickness. (b) Magniﬁ cation of

the transition region. (c) Domain size as a function of Fe thickness. [Reprinted

cal Society.]

682 J. Feng and A. Scholl

that are less than 1 µm in size. The image contrast is the result of the

different angle between the X-ray polarization and the antiferromag-

netic axis in each domain. The ratio image or XMLD image p

B/A

shows

the domains with improved contrast. The spectra were determined

from a sequence of images, an image stack, acquired as a function of

the energy of the incident X-rays. Local spectra were then calculated

within single domains by plotting the local image intensity as a func-

tion of the energy at which the image was taken.

To prove that the perceived image contrast had a magnetic origin,

the antiferromagnetic ordering transition was studied by acquiring

temperature-dependent domain images (Scholl et al., 2000). A reduced

ordering, or Néel, temperature of 670 K was observed in the ﬁ lm, com-

pared to a bulk Néel temperature of 740 K (Figure 9–22). The domain

contrast was compared with a ﬁ t function ∼〈M

〉

, which was estimated

using mean ﬁ eld theory. The good match between ﬁ t and measurement

demonstrated that the antiferromagnetic ordering transition in a thin

ﬁ lm could be described in the mean ﬁ eld approximation.

The exchange coupling and exchange bias at the interface between a

ferromagnet and an antiferromagnet has been attributed to uncompen-

sated spins at the surface of the antiferromagnet (Takano et al., 1997).

This conjecture was not directly proven until surface- and interface-

sensitive X-ray microscopy techniques such as X-PEEM became avail-

able. Figure 9–23 shows magnetic domain images of a Co/NiO(001)

sample, measured using circular polarization (Ohldag et al., 2001; 2003).

The Co ﬁ lm was grown by e-beam evaporation in UHV on a freshly

cleaved and in situ-annealed NiO(001) single crystal. Because of the

surface sensitivity of PEEM, most of the signal in the images originated

from the thin ferromagnet ﬁ lm and the surface region of the antiferro-

magnet. The Ni moment could be separated from the much larger Co

moment, because of the element speciﬁ city of the method. The top

Figure 9–21. X-ray absorption spectra and images of micron-sized antiferro-

magnetic domains in an LaFeO

thin ﬁ lm. The spectra were measured in single

domains and show X-ray magnetic linear dichroism at the L resonances. The

images were acquired at energies labeled A and B. Image p

B/A

is the ratio image.

Chapter 9 Photoemission Electron Microscopy (PEEM) 683

Figure 9–22. Magnetic contrast in LaFeO

as a function of temperature. The

data of the phase transition are compared with a mean ﬁ eld calculation.

AAAS/Science.]

Figure 9–23. Magnetic domain images of Co (top layer, ferromagnetic), NiO

(bottom layer, antiferromagnetic), and NiO interface layer (center layer,

ferromagnetic).

684 J. Feng and A. Scholl

image was acquired using circular polarization at the Co edge and

shows ferromagnetic domains, which were aligned domain-by-domain

with NiO antiferromagnetic domains (bottom image). The center image

was acquired at the Ni edge using circular polarization and shows the

small ferromagnetic moment of the Ni interface layer, which mediated

the coupling between the NiO antiferomagnet and the Co ferromagnet.

The moment was estimated to be close to the moment of one monolayer.

Spectroscopy studies of the chemistry of the interface layer showed that

the uncompensated spins were a result of a chemical reaction at the

interface, which led to a partial oxidization of the Co and a reduction

of the surface of the NiO.

111

Similar results were obtained on Co/LaFeO

/SrTiO

(001) (Nolting et

al., 2000). The interface coupling led to a uniaxial interface anisotropy

and a correlation of the antiferromagnetic and ferromagnetic domains.

Furthermore, a unidirectional anisotropy or bias was observed locally in

single Co domains. This local bias averaged out to zero over a large area

but reached signiﬁ cant values in single, small domains. Remanent mag-

netization loops of single domains showed a different switching ﬁ eld

for opposite directions of the applied ﬁ eld pulse (Figure 9–24). The

Figure 9–24. Magnetization reversal of Co/LaFeO

imaged using XMCD. A

local bias exists in single domains (hysteresis loops a and b). Integrated over

a larger sample area the bias vanishes. [Reprinted with permission from

Nolting et al. (2000).

100