Vij D.R. Handbook of Applied Solid State Spectroscopy

Подождите немного. Документ загружается.

14.5 Applications

The finite attenuation length makes soft X-ray fluorescence spectroscopy a

tempting choice for studying buried interfaces and modifications of solid

surfaces and thin films in a gas atmosphere, where conventional surface

techniques, such as X-ray photoemission spectroscopy, ultraviolet

photoemission spectroscopy, and Auger electron spectroscopy are not usable.

The bulk-probing property of X-ray spectroscopy is often a very useful asset

in materials research, as quite often the decisive structure for a certain

property lies buried in an interface or below a capping layer. In the study of

Si(100) buried in GaAs mentioned earlier, it was demonstrated that it is

possible to extract detailed information about DOS distributions [35]. Ab

initio calculations [36, 37] showed very good agreement with the

experimental spectra and allowed for a quantitative analysis. The layer

thickness and the distribution over atomic sites were concluded to be very

important for the total spectra.

A fundamental research area of current interest concerns the lattice-

mismatched Si/Ge system. Depending on growth conditions, Ge islands or

wetting layers are formed on the Si (100) substrate [38, 39, 40]. Through soft

X-ray emission spectroscopy combined with ab initio density functional

lowest excitation energy (398.5 eV), presumably corresponding to the 1s WR



transition of pyridine-like N, the main emission line is centered at 394.5 eV.

This peak cDQEHDWWULEXWHGWRD

to 1s transitions while the 392-eV shoulder

is mainly due to σ to 1s. This emission pattern is similar to the calculated

spectra for N2, which supports the assignment to pyridine-like N.

When the excitation energy is increased to 399.5 eV, mainly nitrogen in

the N1 structure should be excited. However, the emission spectra indicate

that also a large fraction of the pyridine-like N atoms are excited at this

energy. This can be explained by a relatively broad signal corresponding to

pyridine-like N due to shifts depending on the second nearest neighbors.

Thus, the emission spectrum can be modeled by a superposition of the N1 and

N2 spectra. The excitation energy of 400.8 eV, however, is just below the

excitation threshold of the N3 structure and the second peak of N2, and

therefore mainly the N1 structure is excited, which is also evident from the

sharp emission spectrum corresponding to the nitriles.

The calculated spectrum of graphite-like N (N3) resembles well the broad

experimental spectrum for the excitation energy of 401.4 eV, as we would

expect based on the SXES results. The 408.0-eV spectrum corresponds to

electrons excited to the σ* levels, and since, according to the simulated X-ray

absorption spectra, mainly pyridine-like and graphite-like N have intensity in

this region, the spectrum can be modeled well by a sum of the N2 and N3

spectra.

atoms in different bonding environments are excited depending on the photon

energy is clearly reflected by the differences in the SXES spectra. For the

613

theory (DFT) calculations, fine details in the partial local density of states of

buried layers can be extracted [41]. The bulk DOS of Si and Ge display three

features, each labeled S

, S

,andG

, G

,andG

, respectively. The 1

multilayer (ML) spectrum (“0% mix”) shows six features. It appears that the

state is split into two states G

and G

, and similarly for G

and G

.The

reason for the splitting is the changed symmetry in the Ge layer. Bulk Ge has

a larger lattice constant (5.58 Å) than bulk Si (5.39 Å) and the epitaxial Ge

monolayer is therefore compressed within its plane. As there is a tendency to

keep the unit volume unchanged the Ge-Si spacing perpendicular to the plane

(z direction) increases by 3.5%, which is somewhat lower than that required

for a constant unit volume. The monolayer Ge atoms are thus surrounded by a

tetragonal structure, instead of the bulk cubic structure. The s-states are,

however, not directly influenced by this effect. On the other hand, the three p-

levels, degenerate in cubic symmetry, are split, creating two degenerate levels

directed within the monolayer plane (x,y) and one level directed perpendicular

(z) to the monolayer plane. As we have a mixing of s and p-states in the solid,

the crystal field splitting of the p-states will be reflected indirectly also in the

s-states.

To test this geometrical argument the (s + d) DOS of the buried Ge layer

was recalculated, but with the Si atoms substituted by Ge atoms, keeping all

the previous distances unchanged. Indeed it was found, as summarized in

Figure 14.11, that the low-lying peak G

in bulk Ge splits into two peaks G

’

and G

’

in the Ge monolayer. The peak G

is similarly split. This shows that

the splitting of the levels in the Ge layer in bulk Si is a geometrical effect. A

further shift of the levels occurs due to hybridization between the Ge and Si

states, as illustrated in Figure 14.11.

In the real system the Ge atoms will not form a perfect smooth monolayer,

but due to intermixing the Ge atoms will, to some extent, occur on

neighboring Si sites, and this situation has been simulated by “25% mix.” An

entirely new, complicated structure appears, however, as compared to “0%

mix.” We note, though, that two new strong peaks appear in the 11–12 eV

region (I

and I

), i.e., below any structure in the “0% mix” spectrum.

The agreement between experiment and theory is very good, in particular

for the case of 1 ML Ge. There is a strong indication that the layers are of

high perfection, but that some intermixing with the neighboring Si atoms

occurs. The electronic states in the Ge layer can be viewed as the Si bulk

states tunneling through the Ge layer and resonating on the Ge atoms. Thus

the situation resembles the “virtual crystal approximation” for alloys, where

an averaged electronic structure is observed. Due to the reduced symmetry in

the Ge layer two Ge resonances are observed for each Si bulk state.

14. Soft X-Ray Emission and Resonant Inelastic X-Ray Scattering Spectroscopy 614

14.5 Applications

Figure 14.11 Comparison between experimental SXE spectra and (s + d) DOS theory for 1 ML

Ge in Si (100). The experimental spectrum has been convoluted by 0.3 eV Gaussian.

probe in the characterization of thin film materials. Some examples of the

studies are presented to show the chemical sensitivity, site selectivity, and

depth-dependence of this method. The finite penetration of X-rays offers true

bulk-probing and facilitates studies of interfaces and buried structures.

14.5.2 Nano Structures

The properties of matter at nanoscale dimensions can be dramatically

different from the bulk or the constituent molecules. The differences arise

through quantum confinement, altered thermodynamics, or changed chemical

reactivity. In general, electronic structure ultimately determines the properties

of matter, and it is therefore natural to anticipate that a complete

understanding of the electronic structure of nanoscale systems will lead to

progress in nanoscience and bioscience, not inferior to the progress we have

seen in other fields in recent years.

In short, soft X-ray emission spectroscopy has been used as a chemical

615

In the nano-regime, two effects dominate the chemical and physical

properties: (i) Increasing contribution of surface atoms (surface effect):

decreases. For a 30-nm particle, about 5% of all the atoms are on the surface,

while for 3-nm particle, up to 50% are surface atoms. (ii) Quantum

confinement as the nanoclusters approach a few nanometers (size effect),

when electrons will begin to feel the effects of quantum confinement as the

size of the structure becomes comparable to the electron wavelengths.

The energy position of the band edge can be controlled by the

electronegativity of the dopants, solution pH (flatband potential variation of

60 mV per pH unit), as well as by quantum confinement effects. Accordingly,

band edges and bandgap can be tailored to achieve specific electronic, optical,

or photocatalytic properties. A very important application is found in the

generation of H

from direct photo-oxidation of water without external bias

[42]. Quantum confinement effects on bandgap profiling in similar arrays has

been studied by resonant inelastic X-ray scattering of synchrotron radiation

for potential application of nanomaterials for direct photo-oxidation of water

by solar irradiation.

The measurements were done on synthetic α-Fe

nanorods grown by

controlled aqueous chemical growth. The samples investigated are thin films

that consist of crystalline arrays of hematite nanorod bundles of 50-nm

diameter and 500-nm length perpendicularly oriented onto the substrate. Each

bundle was found to consist of self-assembled nanorods 3–5 nm in diameter

[43]. The samples were prepared by heteronucleation growth and

onto the substrate and subsequently heated in air to 550

&WRDOORZWKHFU\VWDO

phase transition to hematite (α-Fe

) as confirmed by RIXS.

measurements of powder [44] and single-crystal [45] α-Fe

. The spectrum

shows the spin-orbit interaction of the 2p core level that splits the L

(2p

1/2

)

and L

(2p

3/2

that cause multiplets within the edges. The ligand field splitting of 3d

interactions (1–2 eV), gives a 1.4-eV energy splitting between t

(xy, yz, xz)

22 22

spectral shape [46]. First, it splits d orbitals further than the atomic ones by

the formation of molecular orbitals. Second, charge transfer gives rise to the

(in the region of 711–718 eV).

The resonant RIXS spectra at the Fe L-edge of α-Fe

nanorods were

measured and are shown in the bottom of Figure 14.12. Several energy loss

14. Soft X-Ray Emission and Resonant Inelastic X-Ray Scattering Spectroscopy

Surface atom-to-bulk atom ratio increases dramatically as the particle size

Fig. 14.12 (upper). The spectral shape is very similar to the previous XAS

The Fe L-edge absorption spectrum of α-Fe O nanorods is shown in

) edges, and the p-d and d-d Coulomb and exchange interactions

transition metal, being of the same order of magnitude as p-d and d-d

and e (x–y, 3 z–r) orbitals. The charge transfer has two main effects on

tail shape that can be seen on the higher energy side of the edges, especially at

616

thermodynamic stabilization of akaganeite (α-FeOOH) in solution at 90 &

14.5 Applications

d and charge-transfer excitations, are identified in the region from 1 to 5 eV.

2p and iron 3d orbitals. The 2.5-eV excitation, which corresponds to the

bandgap transition of hematite, appears significantly blue-shifted as compared

to the reported 1.9–2.2 eV bandgap of single-crystal and polycrystalline

samples. This finding indicates that such designed nanomaterials would meet

the bandgap [47] requirement for the photocatalytic oxidation of water

without an applied bias.

Figure 14.12 Fe L-edge absorption spectrum of α-Fe

nanorods (upper); energy-dependent

RIXS at Fe L-edge of α-Fe

nanorods (lower).

ZnO, a wide bandgap semiconductor, has attracted considerable attention

during the last years due to its potential technological applications [48, 49, 50,

51, 52] such as, for instance, high efficient vacuum fluorescent displays

(VFD) and field-emission displays. ZnO has also been used for short

wavelength laser devices.

features are clearly resolved. The low energy excitations, such as the strong d-d

The 1-eV energy loss features originate from multiple excitation transitions.

The 4.1 and 6.4 eV excitations originate from charge-transfer between oxygen

617

Intensity (arb. units)

730725720715710705700

Energy (eV)

Fe 2p -absorption

Intensity (arb. units)

-12 -10 -8 -6 -4 -2 0 2 4

Ener

(eV)

RIXS

4.1

2.5

1.0

707.8 eV

709.2 eV

Excitations

6.4

the valence band maximum and conduction band minimum. The absorption-

emission spectrum yields the fundamental bandgap energy of 3.3 eV.

Figure 14.13 Oxygen X-ray absorption and emission spectra reflecting, respectively, the

conduction band and valence band near the Fermi-level of ZnO nanoparticles in comparison

with bulk ZnO.

In order to emphasize the differences of spectra, the spectrum of bulk ZnO

has been subtracted from the spectrum of nanostructured ZnO by using a

weighting factor, as shown in Fig. 14.14. The weighting factor was chosen to

avoid negative intensity, obtained when subtracting the maximum bulk

contribution. The difference spectrum was then multiplied by 3. The feature C

has pronounced intensity compared to features A and B, which can be

correlated to enhanced p-d hybridization in nanostructured ZnO.

14. Soft X-Ray Emission and Resonant Inelastic X-Ray Scattering Spectroscopy

(valence band), and the oxygen K absorption spectrum reflects the O 2p

4s states. In the region of 539–550 eV the spectrum is mainly attributed to O

2p hybridized with Zn 4p states. Above 550 eV, the contribution is mainly

coming from O 2p-Zn 4d mixed states. Stronger s-p-d hybridization was

revealed in nanostructured ZnO since the contributions of features at 520 eV

and 523 eV are enhanced. A well-defined bandgap can be observed between

the X-ray absorption can be mainly assigned to O 2p hybridized with Zn

unoccupied states (conduction band). In the photon energy region of 530–539 eV,

together with the corresponding X-ray absorption spectrum in Figure 14.13

[53]. The oxygen K emission spectrum reflects the oxygen 2p occupied states

The X-ray emission spectra of bulk and nanostructured ZnO are displayed

618

Intensity (arb. units)

545540535530525520515

Ener

(eV )

Bandgap

ZnO

nanocrystalline

single crystal

valence

band

conduction

band

14.5 Applications

Figure 14.14 The difference spectrum of bulk and nanoparticles ZnO, from the X-ray emission

spectra in previous figure.

14.5.3 Transition Metal Systems

14.5.3.1 Cu 2p- and 3p-RIXS of Cuprates

For correlated 3d-transition metal oxides, resonantly excited X-ray emission

spectra. As was pointed out previously the inelastic scattering cross-section

becomes very large at core level absorption resonances. Also, since the

scattering process conserves energy locally the spectra are not subject to

intermediate state lifetime broadening, which otherwise can be a limiting

factor for the spectral resolution. The excited electron that is expelled from

the core level remains close to the atom in a bound state and influences the

spectrum. Local energy conservation entails that the outgoing photon energy

either corresponds to a specific atomic-like excitation (excitonic spectrum)

within the same sub-shell, i.e., a dd-excitation, or more molecular-like

configuration, possible final states for excitation at a p-resonance include

excited| 3d

〉-states and |3d

〉-states (where L

denotes a missing electron in a

effects, therefore, these excitations are also called crystal field excitations.

For cuprates dd-excitations involve the rearrangement of merely one

empty orbital (in octahedral symmetry grouped into a triplet of e

-orbitals: xy,

xz, yz, and a doublet of t

-orbitals:

–y

– z

) and

thus only 9 possible

final states exist. In practice, the narrow spacing between the orbitals only

allows one to resolve two or three RIXS peaks separated by some tenths of an

spectra differ significantly from high energy excited (fluorescence-like)

619

ligand atom). These states are separated in energy (typically in the 1-eV

excitations that involve electron-transfer from the excited atom to a ligand,

range) by the crystal field that is due to electrostatic fields and hybridization

e.g., an oxygen atom. In particular, for cuprates that have a |3d 〉 ground state

Intensity (arb. units)

530528526524522520518516

Ener

(eV )

of ZnO

nanocrystalline

single crystal

difference (Nano-0.85SC)

difference X3

spectrometers offer an ultimate resolution of about one part in several

thousand. Therefore it is difficult to resolve many of the features related to

these low energy excitations at the transition metal L-resonances (metal 2p-

the energy resolution is about 0.5 eV at best. Instead, instrumental resolution

better than 0.1 eV is readily obtained with excitation at the M-resonances

(metal 3p-excitation), which are located below 75 eV. A caveat is the high

cross-section for elastic scattering of low energy X-rays that leads to an

enhancement of the tails of the monochromator function so that weak inelastic

scattering features less than about 0.5 eV below the excitation energy can be

obscured.

The scattering nature of the RIXS process entails that binding anisotropies

lead to angular radiation anisotropy. This in turn can be recorded by varying

the angle between the incident X-ray; polarization vector and the direction of

excitation multiplet was investigated in a pioneering work on the Cu 2p-RIXS

of La

CuO

[54, 55]. Although the spectral resolution was not better than 2

eV, the polarization analysis enabled resolving spectral shifts due to the

dependence of transition probabilities to dd-excitations on orbital symmetry.

Following this example, other RIXS experiments [56] have taken advantage

of the extra information that can be extracted by comparing spectra taken in

the so-called “depolarized” geometry (corresponding to the “vertical position”

in Figure 14.15 versus in the

“horizontal position”).

Figure 14.15 Experimental geometries for RIXS measurements. Linearly polarized

monochromatic synchrotron radiation with the electric field vector in the horizontal plan is

directed towards the sample. The detection direction can be varied continuously between the

horizontal (so-called “depolarized”) and the vertical (so-called “polarized”) spectrometer

positions. A manipulator can orient the sample axes in any desired direction.

14. Soft X-Ray Emission and Resonant Inelastic X-Ray Scattering Spectroscopy

excitation), which are located in the approximate range 400

1000 eV, where

the detected X-rays; see Figure 14.15. The polarization-dependence of the dd-

620

polarized” geometry (corresponding to the

“

electron volt. This brings us to the question of instrumental resolution: present

14.5 Applications

As pointed out above, energy resolution is much higher at lower energies.

Cu 2p-RIXS requires excitation energies above 900 eV compared to 75 eV for

Cu 3p-RIXS. Figure 14.16 shows a comparison of Cu 2p-RIXS and Cu 3p-

RIXS for the cuprate CuGeO

using the best instrumental resolution available

[57]. The peaks at about 1.6 eV are crystal field excitations that have

splittings of a few tenths of an electron volt. Crystal field excitations have

been studied for a long time at very high energy resolution in infrared

spectroscopy or Raman scattering spectroscopy. However, interpretation and

calculations of optical spectra are much more intricate than for RIXS (which

in analogy has also been called resonant Raman X-ray scattering). RIXS is a

dipole allowed two-photon (i.e., second order) scattering process described by

the Kramers-Heisenberg formula connecting, in the first step, the ground state

Figure 14.16 Cu 2p-RIXS (solid line) and Cu 3p- RIXS (dashed line) of CuGeO

shown on an

energy loss scale.

with a core hole-excited, charge-neutral excitation (intermediate state) and, in

the second step, connecting the intermediate state with a charge-neutral,

621

Normalized intensity (arb.units)

-10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2

Energy loss (eV)

Cu 3p-RIXS

(resolution 80 meV)

Cu 2p-RIXS

(resolution 0.7 eV)

charge transfer

excitations

CuGeO

The first demonstration of 3p-RIXS of a cuprate was performed on

CuO

by Kuiper et al. [59]. The lifetime broadening of the 3p core hole

state of approximately 1.5 eV notwithstanding, features separated by a few

tenths of an eV could be resolved, as noted in a previous section. In particular,

this study addressed the controversy [60] whether one has found dd-

excitations at sufficiently low energies to have relevance in the mechanism for

high-T

superconductivity. This would require that excitations as low as 0.5

eV or below exist, something for which some groups claim evidence exists

from optical Raman spectroscopy. However, no excitation below 1.5 eV is

found in Cu 3p-RIXS and therefore it is likely that dd-excitations at such low

energies do not exist.

Also, other types of excitations appear in Cu p-RIXS, e.g., charge transfer

(CT) excitations that leave the atom in a |L

〉 final state. RIXS

measurements of CT-excitation energies in connection with model

calculations are highly interesting with respect to magnetic interactions

because the CT energy is the dominating term of the superexchange

mechanism parameter J. CT states require higher excitation energies and are

spread out reflecting the oxygen 2p bandwidth: CT states are seen in Figure

14.16 as a weak band between about 4 eV and 8 eV energy loss. Since CT

states have a valence hole in the oxygen ligand it is expected that these

excitations are much stronger in O 1s-RIXS (see next Section,

14.5.3.2).

14.5.3.2 Oxygen 1s RIXS of Cuprates

The electronic structure of cuprates is determined by the interplay of the

narrow Cu 3d band localized on the Cu sites and the predominantly itinerant

O 2p band. The bonding of the Cu-O octahedra themselves and their coupling,

e.g., edge- or vertex-sharing plates or chains carry key information on the type

of macroscopic behavior of these compounds, be it

superconductivity or, as in the case of CuGeO

, a spin-Peierls state (TSP = 14

K). Although a vast amount of literature has been dedicated to the

understanding of high-T

superconductivity, the underlying mechanism is still

not fully understood. Spin-Peierls states are unusual for inorganic materials

and it is interesting to note that the one-dimensional edge-sharing CuO

14. Soft X-Ray Emission and Resonant Inelastic X-Ray Scattering Spectroscopy

low temperature

622

valence-excited state obeying the selection rule ∆l = 0, ± 2 (which includes,

be found in the review by Kotani and Shin [58].

e.g., the ground state and dd-excitations). More on theoretical calculations can