Koster G., Rijnders G. (Eds.) In situ Characterization of Thin Film Growth

Подождите немного. Документ загружается.

62 In situ characterization of thin ﬁlm growth

3.2.2 Instrumentation and light sources for high-

resolution photoemission spectroscopy

ARPES experienced a renaissance in the late 1990s and early 2000s with

the advent of 2D multiplexing analyzers by the Swedish company VG

Scienta which could measure a range of electron kinetic energies and angles

simultaneously. The way that traditional electron analyzers worked was that

their angular acquisition was based on a small pinhole aperture. The position

of the pinhole relative to the sample normal determined the photoelectron K,

and the solid angle subtended by the pinhole determined Dk. The problem

with this approach is that all electrons outside this pinhole are discarded,

and only one wavevector can be measured at a time. The great advantage

of the Scienta analyzers is that the electron lens system for these analyzers

allowed for parallel detection of many (~100) angular channels, thereby

increasing the measurement efciency by about two orders of magnitude.

Today, such multiplexing analyzers are quite common and easily achieve

energy resolutions of DE = 1 meV and simultaneously measure angular

ranges greater than 30 degrees in parallel. The electrons pass through the

entrance aperature, and then through a three-element lens system, through

the entrance slits, and then the two concentric hemispheres housed inside

the large stainless steel dome (all electron lens elements are electrostatic).

The electrons are nally detected at the bottom of the analyzer, after the

electron signals are amplied by a pair of microchannel plates (MCPs) and

a phosphor screen. Flashes on the phosphor screen are then detected ex situ

by a CCD camera looking through the window, visible at the bottom ange.

An illustration of a Scienta analyzer operating at a synchrotron beamline is

shown in Fig. 3.4.

A major barrier to performing UPS is the relative dearth of appropriate UV

and VUV photon sources. As a result, many UPS and ARPES systems are

Detector

Sample

Toroidal

mirror

Exit slit

Scan

Spherical

mirror

Four-jaw

aperture

Plane

mirror

Electrostatic lens

Entrance slit

Scienta

hemispherical

analyzer

Undulator

Plane

gratings

3.4 Schematic of Scienta analyzer on a beamline. Photoelectrons

with different take-off angles are imaged onto the long entrance

slit, and pass through the concentric hemispheres, and strike the

microchannel plates. From Damascelli et al. (2003).



63Ultraviolet photoemission spectroscopy (UPS)

located at synchrotrons where at undulator insertion devices, tunable photon

energies can be achieved with high photon uxes (~ 10

photons/s mm

)

within a narrow (~ meV) energy bandwidth. Synchrotron sources remain

the light source of choice for ArPeS, as no other sources can match their

exibility in terms of photon energies and polarizations. However, because

of the scarcity and relatively small number of synchrotrons, laboratory-based

UV and VUV sources present a highly desirable alternative. These sources

must exceed the energy of the work function (typically 4–5 eV), and such light

sources are relatively scarce. The most common light sources for laboratory-

based UPS are plasma discharge (He, Ar, Xe) lamps which have discrete

emission lines with high brightness (~ 10

/s mm

) and narrow spectral widths

(~ 1 meV). The main disadvantage to plasma discharge lamps is the lack of

tunability in photon energies which can be essential for many experiments.

More recently, VUV lasers which use higher harmonic generation (HHG)

have come to the forefront. Although currently restricted in photon energy

tunability and limited to fairly low photon energies (5–7 eV), they have the

distinct advantage of superior spectral linewidths and brightness over plasma

discharge lamps. In addition, laser sources can be utilized for time-resolved

experiments such as pump-probe measurements. However, at present, the

wavelengths available to VUV lasers is rather restricted, although this is

likely to change as the technology matures.

3.3 Applications of UPS to thin ﬁlm systems

3.3.1 SrRuO

SrRuO

has drawn wide interest for thin lm studies and applications for

a number of major reasons. First, SrRuO

is a ferromagnetic metal (T

160 K) which is unusual for a 4d transition metal oxide, but also exhibits

unconventional ‘bad metal’ behavior at elevated temperatures. The Ruddlesden–

Popper series of layered ruthenates Sr

n+1

3n+1

, of which SrRuO

is the

three-dimensional end-member, display a remarkable evolution in magnetic

and electronic properties as a function of dimensionality, n. The single layer n

= 1 compound, Sr

RuO

, is an exotic spin triplet superconductor with a time-

reversal symmetry breaking ground state (Mackenzie and Maeno 2003). The

n = 2 bilayer compound, Sr

exhibits quantum critical metamagnetism

and an apparent electronic nematicity where the magnetoresistance breaks

C4 symmetry (Borzi et al. 2007). Ferromagnetic tendencies are enhanced

by increasing the number n of RuO

sheets per unit cell. In addition, Ca

substitution for Sr dramatically alters the magnetic and electronic properties

of all known members of the Sr

n+1

3n+1

ruthenates, despite the fact that

Ca and Sr are isovalent (for instance, CaRuO

is a paramagnetic metal, and

RuO

is an antiferromagnetic insulator). In addition, spin–orbit interactions



64 In situ characterization of thin ﬁlm growth

and 4d orbital degeneracies are known to be important factors. Therefore,

SrRuO

and other members of the ruthenate family present a wealth of exotic

electronic and magnetic behaviors which are ripe for investigation. From the

thin lm standpoint, SrRuO

is also an excellent material for investigation

since epitaxial thin lms of very high quality can be synthesized. Until very

recently, many properties of lms (such as resistivity) were found to exceed

even those of the best bulk single crystals. In addition, the low resistivity of

SrRuO

lms make it an excellent candidate material for thin lm applications

such as electrode materials or magnetic tunnel junctions. In addition, the

pseudocubic perovskite structure makes SrRuO

suitably lattice matched to

other common transition metal oxides such as SrTiO

, a common substrate

material for SrRuO

Because the electronic properties of SrRuO

are still not well understood,

UPS and ARPES studies of SrRuO

are essential for answering important

basic questions about the electronic structure, such as whether rst-principles

calculations can capture an accurate description of this complex correlated

material which exhibits ferromagnetism and three-fold orbital degeneracy, or

the nature of the large effective masses and the strength of electron–electron

correlations. However, because of its pseudocubic structure, cleaving single

crystals does not yield the at, well-dened surfaces necessary for ARPES.

Therefore, measurements of atomically at thin lm samples synthesized in

situ are essential for performing ARPES measurements of the quasiparticle

band structure and Fermi surface. Moreover, because of the surface sensitivity

of the ARPES technique (inelastic mean free path ~ 1 nm), these thin lm

samples cannot be removed from ultra-high vacuum (UHV) conditions before

measurement; exposure of an atomically pristine surface to atmosphere can

render a surface unmeasurable within nanoseconds. Therefore, the synthesis,

transfer, and measurement stages must all take place under strict UHV

conditions.

The earliest UPS measurements of SrRuO

were performed by Fujioka et

al. (1997) and Okamoto et al. (1999) on polycrystalline samples of SrRuO

which were sintered as pellets and scraped in situ. Based on comparisons to

the density of states obtained from band structure calculations, Okamoto et

al. were able to obtain a mass enhancement factor of approximately m*/m

4.4, which appeared consistent with estimates from transport measurements.

On the other hand, based on this analysis, it is difcult to determine the origin

of this fairly large mass enhancement, although it has been speculated by

Fujioka et al. that this may be due to electron–electron interactions. On the

other hand, Mazin and Singh (1997) have argued that electron–magnon and

phonon couplings in SrRuO

could be anomalously large due to the important

role of oxygen in forming the ferromagnetic state. A major limitation of

performing photoemission spectroscopy on polycrystalline samples is the lack

of any momentum-resolved information. Therefore, developing a method for



65Ultraviolet photoemission spectroscopy (UPS)

pursuing in situ photoemission studies of SrRuO

thin lms will be critical

for in-depth studies of the electronic structure of SrRuO

Using a novel combination of both PLD and MBE, Siemons et al.

(2007) were able to synthesize a variety of SrRuO

thin lms with varying

degrees of Ru stoichiometry and quantied the strong dependence of the Ru

stoichiometry (deciency) on the unit cell volume and c-axis lattice constant

as well as with the residual resistivity ratio (RRR), as shown in Fig. 3.5. They

Stoichiometric ﬁlm (MBE) 2q = 45.96°

Ru poor ﬁlm (PLD) 2q = 45.90°

Both ﬁlms are approximately 330 nm thick

(220) SrRuO

(002) SrTiO

45.50 45.75 46.00 46.25 46.50

2q (degrees)

(a)

Log intensity (arb. units)

at 300 K

at 4 K

240.6 240.8 241.0 241.2 241.4 241.6 241.8

Unit cell volume (Å

)

(b)

Residual resistivity ratio (R

300 K

4 K

)

Resistivity (µΩ cm)

100

300

250

200

150

100

3.5 Characterization of the dependence of (a) the c-axis lattice

constant and (b) the resistivity and residual resistivity ratio

300 K

) on the Ru stoichiometry of SrRuO

thin ﬁlms grown by

PLD and MBE, as characterized by Siemons et al. (2007).



66 In situ characterization of thin ﬁlm growth

determine that lms grown by PLD are typically ruthenium decient with

poor (low) residual resisitivities and can have an expanded unit cell volume,

and could synthesize thin lms with RRRs as low as 5 or less (PLD, high

oxygen pressures) or up to 40 (MBE, low oxygen pressures). It is interesting

to note that the Curie temperature is only very weakly dependent on Ru

stoichiometry, and therefore should not be used as a metric for determining

the quality of SrRuO

thin lms. In addition, using in situ photoemission

spectroscopy, Siemons et al. clearly demonstrate that the quality of the UPS

spectra of these thin lms near E

(E < 1 eV) is strongly dependent on the

Ru stoichiometry. In particular, the spectral intensity near E

(constituted

primarily of the ru d

, d

, and d

orbitals) is strongly suppressed with

respect to the background at higher binding energies as shown in Fig. 3.6.

Therefore, measurements which attempt to extract the quasiparticle mass

enhancement based solely on comparisons of spectral weight intensities

and comparisons with theoretical densities of states must necessarily take

into account the potentially sizable effects of Ru non-stoichiometry on the

photoemission spectra.

As mentioned earlier, the electronic and magnetic properties of the

Ruddlesden–Popper series ruthenates are known to depend critically on

dimensionality. In response to this, ultrathin lms of SrRuO

grown on

SrTiO

have been investigated and an evolution in their electronic and

Stoichiometric ﬁlm (MBE)

Ru poor ﬁlm (MBE)

Ru poor ﬁlm (PLD)

Ru t

Incoherent Coherent

2.5 2.0 1.5 1.0 0.5 0.0 –0.5

Binding energy (eV)

Counts (arb. units)

3.6 Dependence of the near E

spectral weight on the Ru

stoichiometry in thin ﬁlms of SrRuO

as measured by in situ UPS

using a He I discharge lamp. From Siemons et al. (2007).



67Ultraviolet photoemission spectroscopy (UPS)

magnetic properties have been reported by increasing the SrRuO

lm

thickness within the rst 10 monolayers (MLs) of growth. Xia et al. (2009)

have determined through a combination of polar Kerr effect measurements

and resistivity that below a SrRuO

lm thickness of 4 MLs, the material’s

ferromagnetic properties appear to be heavily suppressed, suggesting the

formation of antiferromagnetism, while above 4 MLs, bulk-like behavior is

reported. In conjunction with this ferromagnetic suppression, lms thinner

than 4 MLs exhibited insulating behavior in resistivity measurements, shown

in Fig. 3.7. These ndings are consistent with earlier reports by Toyota et al.

(2005) which studied the evolution of the photoemission spectra of SrRuO

lms as a function of thickness. As shown in Fig. 3.8, below approximately

6 MLs, there is little to no spectral intensity near E

, while above 10 MLs

the spectra are similar to much thicker (100 ML) lms. This again points to

some kind of metal–insulator transition around 4 MLs which also appears

to be coincident with a dramatic change in the ferromagnetic properties of

the material.

The work by Siemons and Toyota clearly establishes the importance of in

dr/dT

–5¥10

–1¥10

–1.5¥10

–2¥10

–2.5¥10

–1

–2

–3

–4

–0.1

–0.3

–0.5

–0.7

–0.9

4 ML

3 ML

Thick ﬁlm

0 10 20 30 40 50

T (K)

0 50 100 150 200 250

T (K)

0 50 100 150 200 250

T (K)

2 ML

3 ML

4 ML

6 ML

9 ML

Thick ﬁlm

Resistivity (µΩ cm)

0 50 100 150 200 250 300

Temperature (K)

1¥10

1000

3.7 Resistivity of SrRuO

ﬁlms showing the dependence of the

resistivity and the ferromagnetic transition on the ﬁlm thickness in

monolayer (ML). From Xia et al. (2009).



68 In situ characterization of thin ﬁlm growth

situ photoemission spectroscopy on uncovering the complex electronic structure

of SrRuO

. However, there is still much to be learned about the underlying

electronic structure which may come from future angle-resolved measurements

as well as additional thickness dependence studies of ultrathin SrRuO

lms.

At present, angle-resolved photoemission measurements of SrRuO

have not

yet been conducted, although these recent successes of in situ angle-integrated

photoemission measurements on SrRuO

and in situ ArPeS on other thin

lms such as La

1–x

MnO

suggest that future ArPeS measurements on

SrRuO

lms are not far off on the horizon. ARPES measurements will be

Intensity (arb. units)

SrRuO

ﬁlm

SrRuO

ﬁlm

hn = 600 eV

100 ML

50 ML

20 ML

17 ML

15 ML

12 ML

10 ML

8 ML

6 ML

5 ML

4 ML

3 ML

2 ML

1 ML

STO

100 ML

50 ML

20 ML

17 ML

15 ML

12 ML

10 ML

8 ML

6 ML

5 ML

4 ML

3 ML

2 ML

1 ML

STO

12 8 4 E

3 2 1 E

–1

Binding energy (eV) Binding energy (eV)

(a) (b)

3.8 In situ photoemission valence band spectra of SrRuO

grown

by PLD as a function of nominal ﬁlm thickness, showing depleted

spectral weight near E

below 5 ML. From Toyota et al. (2005).



69Ultraviolet photoemission spectroscopy (UPS)

key in answering critical questions about the intrinsic electronic structure of

SrRuO

, such as whether density functional theory can adequately predict the

electronic structure of such a complex, correlated multi-band ferromagnet,

or what is the true nature (electron–electron or electron–boson coupling)

of the large mass renormalization observed in SrRuO

. However, a major

consideration will be that accounting for Ru stoichiometry will be particularly

critical for high-quality ARPES measurements. A high concentration of Ru

vacancies will introduce a large amount of scattering which would broaden

and mix otherwise well-dened eigenstates in momentum space, meaning

the materials requirements for ARPES will be even more stringent than for

angle-integrated UPS.

3.3.2 La

1–x

MnO

The re

1–x

MnO

manganites (with Re a rare earth, and Ae an alkaline

earth metal) have been at the forefront of the eld of strongly correlated

electrons due to their host of electronic phases and dramatic phase transitions.

Most well known amongst these properties is colossal magnetoresistance

(CMR), where the electrical resistance can drop by more than six orders of

magnitude with the application of a magnetic eld. La

1–x

MnO

(LSMO)

shows excellent metallic behavior and the highest Curie temperature of

all re

1–x

MnO

variants, and is predicted theoretically to be 100% spin

polarized. Based on these properties, LSMO is a potentially ideal material for

the development of new spintronics devices, particularly magnetic memory

and magnetic sensors. Toward this end, CMR effects in LSMO tunnel

junctions and the integration of semiconductors into LSMO-based devices

have been demonstrated, and a great deal of interest and work has been

devoted to synthesizing epitaxial thin lms of LSMO and other manganite

compounds. Recently, there has also been great activity in using LSMO as a

building block for novel heterostructured materials. Cation-ordered versions

of LSMO, where layers of LaMnO

are alternately stacked with layers of

SrMnO

, have shown new and dramatic metal–insulator transitions, increased

ordering temperatures, and may generate spin-polarized two-dimensional

electron gases at the LaMnO

–SrMnO

interfaces.

Due to its three-dimensional pseudocubic structure, ARPES measurements

of bulk single crystals of LSMO remain inaccessible because of the inability

to produce a well-dened crystallographic surface upon cleaving. Therefore,

like the case of SrRuO

, atomically at thin lm samples of LSMO grown

in situ using MBE or PLD provide the only route to accessing the electronic

structure of these materials. The importance of understanding the complex

electronic structure of this strongly correlated material has led to various

research groups performing in situ ARPES studies of thin lms of LSMO.

The rst in situ ARPES studies of thin lms of LSMO were reported



70 In situ characterization of thin ﬁlm growth

by Shi et al. (2004) and later by Chikamatsu et al. (2006). Thin lms were

synthesized using pulsed laser deposition on SrTiO

substrates. Measurements

were conducted using an energy resolution of 150 meV and an angular

resolution of 0.5° for the work of Chikamatsu et al. and 40 meV and 0.2°

for Shi et al. Chikamatsu et al. demonstrated for x = 0.4 that the valence

band (within ~ 10 eV of the Fermi level) exhibits generally good agreement

with band structure calculations based on a local density approximation

plus on-site Coulomb interactions (LDA + U), shown in Fig. 3.9. Close to

(within 1 eV), only broader features with weak spectral weight could

be observed near E

. While the approximate dispersion of these features

qualitatively tracked some of the LDA + U predictions, important questions

persist about the nature of these near E

states. First, the reason for the

dramatically reduced near E

spectral weight is still unclear, although it has

been suggested that this spectral weight suppression could be the signature

of polaronic behavior due to strong electron–phonon interactions, shown in

Fig. 3.10. However, it may be conceptually challenging to reconcile small

Binding energy (eV)

003

103

Wave vector

(a)

Binding energy (eV)

G X G

Wave vector

(b)

3.9 In situ ARPES measurements of the valence band of

1.6

0.4

MnO

thin ﬁlms, showing the (a) second-derivative plots of

the valence band and (b) band structure calculations using the local

density approximation plus on-site Coulomb interactions (LDA + U).

Solid and dashed lines correspond to spin majority and minority

bands respectively. From Chikamatsu et al. (2006).



71Ultraviolet photoemission spectroscopy (UPS)

polaron formation (and subsequently very large effective masses) with its

reasonably metallic behavior at low temperatures (r ~ 10

–5

W cm).

The work by Shi et al. on lms of x = 0.33 grown by PLD also revealed

that the large-scale electronic structure in the valence band also agreed

reasonably well with predictions, but similar to the work by Chikamatsu

et al., the near E

electronic structure remained somewhat surprising. In

particular, in both Chikamatsu’s and Shi’s works, a large sheet of Fermi

0.6

0.4

MnO

/SrTiO

hn = 88 eV T = 25 K

0.6

0.4

MnO

/SrTiO

hn = 88 eV T = 25 K

Intensity (arb. units)

28°

24°

20°

16°

12°

28°

24°

20°

16°

12°

1.0 0.5 E

Binding energy (eV)

(a)

1.0 0.5 E

Binding energy (eV)

(b)

3.10 ARPES measurements of the near E

spectra of La

1.6

0.4

MnO

thin ﬁlms enlarged around the G (k = (0,0,0)) point. The open

symbols represent the approximate energy positions of possible

dispersive features. Panel (a) shows raw data; panel (b) shows data

after subtraction of a high binding energy background. ‘A’ denotes

the dispersive low energy band, while ‘B’ denotes the weakly

dispersive high energy feature. From Chikamatsu et al. (2006).

? ?

