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

Подождите немного. Документ загружается.

1120 U. Weierstall

trons and then decay by the emission of photons (Persson and Baratoff,

1992). Besides these inelastic excitations, photons can also be emitted

due to electron-hole recombination in semiconductors (Abraham et al.,

1990). Given that the tunneling currents used for excitation are rather

small—typically in the nanoampere range—the photon count rates

tend to be small as well. The photon creation efﬁ ciency is typically of

the order of 10

−4

–10

−3

photons per tunneling electron (Persson and

Baratoff, 1992). Increasing the tunneling current in many cases leads

to spontaneous surface modiﬁ cations, which prevents reproducible

experiments. Given the low photon count rates, the photon detector has

to be optimized for maximum collection efﬁ ciency. A number of designs

have been reported, which make use of lenses (Berndt et al., 1991b;

Hoffmann et al., 2002a), mirrors (Berndt et al., 1991b; Nilius et al., 2000),

transparent tips (Smolyaninov et al., 1990; Murashita, 1999), and optical

ﬁ bers (Arafune et al., 2001). The emitted light can be experimentally

investigated using a variety of methods (Reihl et al., 1989):

1. Isochromat spectroscopy: The photon energy is kept ﬁ xed while

the bias voltage is being scanned (at constant tunneling current). This

allows study of the inﬂ uence of the excitation energy on the intensity

of particular emission features. During the acquisition of isochromat

spectra the tip–sample distance is varied by the STM feedback loop to

maintain constant current. The detectors used are avalanche photodi-

odes or photomultipliers, where the spectral response is limited by an

optical bandpass ﬁ lter.

2. Fluorescence spectroscopy: The electron energy (i.e., the bias

voltage) is kept ﬁ xed while the wavelength-resolved distribution of the

emitted photons is measured. This involves a grating spectrometer and

a cooled CCD detector. Direct information on light-emitting transitions

is obtained. This mode allows the observed photon emission to be

assigned to elementary processes such as interface plasmons in the

tip–sample cavity or interband transitions in the sample.

3. Luminescence spectroscopy: A special case of (2) for semiconduc-

tors, where the injected electron thermalizes to the bottom of the con-

duction band emitting a characteristic photon.

4. Spatial mapping: A selected feature in the spectra is used to gen-

erate a high-resolution picture of the intensity of the emitted photons.

Spatial maps are usually acquired simultaneously with a constant-

current image to relate the spectral map to topographic features. In the

case of luminescence this allows the spatial identiﬁ cation of defects,

grain boundaries, and dopant concentrations on semiconductor sur-

faces similar to cathodoluminescence in electron microscopy.

STM-induced photon emission from semiconductors has been com-

pared to cathodoluminescence (CL) (Gustafsson et al., 1998) in scan-

ning electron microscopy, photoluminescence (PL) (Montelius et al.,

1992) and inverse photoemission spectroscopy (IPS) (Reihl et al., 1989).

Measuring luminescence from semiconductors locally by inducing it

with an STM has distinct advantages: Electron tunneling from the tip

provides a bright and extremely localized source of electrons. In addi-

tion and unique to STM measurements, holes can be injected by biasing

Chapter 17 Low-Temperature Scanning Tunneling Microscopy 1121

the tip positive with respect to the sample. Comparative measurements

of STM-induced photon emission in the ﬁ eld emission regime (eV > Φ)

on Si(111) with STS and normal incidence IPS demonstrated a correla-

tion between those methods, but showed also differences due to the

local nature and the strong electrostatic ﬁ eld of the STM measurement

(Reihl et al., 1989).

STM-induced photon emission from metal surfaces involves excita-

tion of a plasmon localized within the region near the end of the tip

(Aizpurua et al., 2000). The plasmon then decays into photons, which

are detected in the far ﬁ eld. These plasmons are excited either by

inelastic tunneling, or by elastic tunneling into the substrate and sub-

sequent thermalization via plasmon generation. Model calculations

favor inelastic tunneling, which occurs in the tunneling gap as the

excitation mechanism (Berndt et al., 1991a). This is in contrast to STM-

induced luminescence on semiconductors, where hot electron decay

within the semiconductor was found to dominate (Abraham et al.,

1990). The resulting photon spectrum resulting from plasmon decay is

quite broad and has a characteristic energy cutoff determined by the

sample bias voltage at eV

bias

. The spectrum is also sensitive to the geom-

etry of the tunnel junction and the tip material (Berndt et al., 1993).

Fluorescence due to hot electron thermalization, on the other hand, is

expected to be insensitive to the tunneling gap conditions and produce

a distinct peak (or a series of peaks) in the photon spectrum. Measure-

ments on Cu(111) in the ﬁ eld emission regime found a close correlation

between oscillations in the conductance, which arise from standing

waves in the tip–sample gap, and oscillations in isochromat photon

spectra (Berndt and Gimzewski, 1993). These ﬁ ndings supported the

view that for metals inelastic tunneling and coupling to a tip-induced

plasmon mode occurs in the tunneling gap.

Low temperature in CL (Murashita et al., 1993), PL, and IPS is known

to increase the intensities of spectral features and/or reduce thermal

broadening, which can obscure important details in the spectrum. This

is one reason to perform STM photoemission measurements at low

temperatures; the other main reason is to improve the long-term drift

stability necessary to record spatial maps (Hoffmann et al., 2004) and

photon spectra from single molecules (Qiu et al., 2003).

Low-temperature luminescence experiments on GaAs/AlAs using

an LT-STM with a transparent tip as detector have been reported

(Murashita, 1997). Isochromat spectra from individual several-atom

silver chains assembled with an LT-STM on the NiAl(110) surface have

been measured showing sensitivity to the number of atoms in the silver

chain (Nazin et al., 2003a). The changes in photon emission with chain

size were explained as a quantum size effect. Recently experiments

with superconducting tips on a superconducting sample were reported

that showed an energy cutoff at 2eV

bias

in the isochromat photon spec-

trum. This could not be explained by single electron tunneling and

was attributed to tunneling of Cooper pairs (Uehara et al., 2001).

The modes of measurement described above (1–4) either provide

lateral or spectral resolution. Spectral and spatial resolution can be

obtained simultaneously by recording a spatial map where at each

1122 U. Weierstall

point in the image a complete ﬂ uorescence spectrum is recorded. The

tip motion during constant-current imaging is delayed at each pixel for

a photon detection time of 0.05–1 s (depending on detector sensitivity).

This mode was named spectroscopic imaging (Hoffmann et al., 2002a).

From the resulting three-dimensional data set isochromat spatial maps

at various wavelengths can be extracted. Moreover, changes of the

sample or tip are easily recognized. The ﬁ rst measurements of this kind

were conducted by Nishitani et al. (1998) at room temperature. The

recording time for a complete spectroscopic image was 1 h. Subsequent

improvements in the photon detection system enabled faster recording

times and the use of an LT-STM provided less thermal drift (Hoffmann

et al., 2002a). Atomically resolved isochromat spatial maps have been

recorded of the Au(110) surface at 4 K where the atomic rows appear

darker then the troughs (see Figure 17–41). The ﬁ nding that atomic scale

structures cause clear variations of the photon emission characteristics

was surprising when it was ﬁ rst observed (Berndt et al., 1995). Theoreti-

cally the local photon emission on metals was expected to extend

laterally over approximately 50 Å

, which is the lateral extent of the tip-

induced plasmons. The ﬁ rst ﬁ nding of atomic scale features in STM-

induced photoemission on Au(110) was explained in terms of the

tip–sample distance dependence of the photon emission, which is given

by the coupling to localized plasmon modes (Berndt et al., 1995). This

model could not explain later measurements that showed that photon

emission is reduced by a factor of 5 at Ag(111) steps, whereas it is

Figure 17–41. Atomic scale detail in photon maps: (a) Constant current image

of a reconstructed Au(110) surface showing two terraces separated by a mono-

atomic step. Marker I: on top of an atomic row. Marker II: below a monoatomic

step. Marker III: between atomic rows. (b and c) Simultaneously acquired

photon map is in registry with the topographic data. (b) Photon wavelength

λ = 671–675 nm. (c) Photon wavelength λ = 762–765 n m. (d) Representative

spectra normalized to the same peak intensity. Inset: Unnormalized spectra.

(From Hoffmann et al., 2004.)

Chapter 17 Low-Temperature Scanning Tunneling Microscopy 1123

increased by 50% at Au(110) steps. In the new study of atomic resolution

spectroscopic imaging on Au(110) mentioned above (Hoffmann et al.,

2004), a new model was presented that is consistent with the existing

data. Atomic resolution on Au(110) and the contrast at steps was attrib-

uted to the local electronic structure and its effect on the elastic and

inelastic tunneling channels (Hoffmann et al., 2004).

The sensitivity of STM-induced photon emission from metals to the

conditions of the tunneling gap has been discussed above. New phe-

nomena are observed when molecules are placed in this cavity. In

spatial photon maps of close packed C

monolayers on Au(110), indi-

vidual molecules have been resolved as distinct maxima (Berndt et al.,

1994). This could not be explained by photon emission from tip-induced

plasmon modes alone, since on metals the emission intensity would be

reduced if the tip–metal distance is increased. Therefore if the emission

from the C

-covered metal surface was directly due to plasmon

modes, the increase in tip–metal distance due to the molecules should

weaken the plasmon modes. But actually the opposite is observed. The

role of the molecules in the emission process is unclear, although

molecular ﬂ uorescence has been suggested.

Spatial photon maps and ﬂ uorescence spectra from single hexa-tert-

butyl-decacyclene (HBDC) molecules on Au(111), Ag(111), and Cu(111)

have been recorded at 4 K (Hoffmann et al., 2002b). Low temperature

was essential for this measurement, since these molecules are mobile

on the surface at room temperature. Results similar to C

are reported,

i.e., increased emission on the molecule, but no new spectral features

attributable to molecular ﬂ uorescence were observed. Only a 4-nm

shift of the HBDC photon spectrum toward shorter wavelengths was

observed compared to the substrate spectrum.

On a metal surface, the electronic levels of a molecule are consider-

ably broadened whereas light emission is strongly quenched (Barnes,

1998), making it difﬁ cult to detect molecule-speciﬁ c emission. In an

LT-STM study it has been shown that molecular ﬂ uorescence could be

observed when the molecule (ZnEtiol) was supported on a thin alumi-

num oxide (Al

) ﬁ lm grown on an NiAl(110) surface (Qiu et al., 2003).

The oxide spacer reduces the interaction between the molecule and the

metal. The photon emission spectra and intensities varied with differ-

ent tips and increased photon emission efﬁ ciency was observed with

Ag tips compared to W tips. This has been attributed to plasmon excita-

tion in the tunneling gap. To reduce the plasmon signature in the

spectra the tips were voltage pulsed. After that, photon emission

spectra with sharp features could be seen when the tip was positioned

directly above a molecule. Furthermore, the spectra were very sensitive

to the tip position above the molecule. Light emission has been found

to be almost identical for molecules of the same conformation. This is

shown in photon spectra form three different molecules (in the same

conformation) measured with three different STM tips (Figure 17–42A).

The spectra could be explained as a superposition of tip-induced

plasmon emission (which depend on tunneling gap conditions) and

molecular ﬂ uorescence (which should be nearly independent of gap

conditions). Molecular ﬂ uorescence is due to electrons tunneling elasti-

1124 U. Weierstall

cally into the molecule and creating excited vibrational states that

make a radiative transition. To eliminate the inﬂ uence of plasmons

from the spectra, they have been divided by the spectra taken from the

NiAl surface (Figure 17–42B) with the same tips. The resulting curves

(Figure 17–42C) show remarkable similarity, indicating that the mole-

cules have nearly identical light emission properties independent of

the tip. The vibrational features were equidistant in energy with

spacing of 40 ± 2 meV (Figure 17–42, inset).

Figure 17–42. Molecular ﬂ uorescence from the excitation of vibrational modes

in a single molecule. (A) Photon emission spectra for three different experi-

ments with three different tips. The spectra where obtained over a single

ZnEtiol molecule shown in the inset (constant current image) at the marked

spot. The raw data are plotted together with the smoothed curves. Spectra 1

and 2 were taken with two different Ag tips and have been offset for clarity.

Spectrum 3 was obtained with a W tip. The differences in the spectra are due

to different tip plasmon properties. (B) NiAl light emission spectra measured

with the same tips and voltages as in (A). (C) Molecular spectra from (A)

divided by the corresponding NiAl spectra from (B) show remarkable similar-

ity. The inset shows the photon energy of each peak determined for the three

spectra. (From Qiu et al., 2003.)

Chapter 17 Low-Temperature Scanning Tunneling Microscopy 1125

A similar picture for the emission process has been used to explain

the photon emission measured with an LT-STM from quantum well

structures formed by Na layers on a Cu(111) surface (Hoffmann et al.,

2001). The photon emission in this experiment was assigned to transi-

tions between the electronic levels of the quantum structure and was

observed together with tip-induced plasmon emission.

Recently another experiment with a room temperature STM has

shown molecular ﬂ uorescence from organic molecules decoupled from

an Au(100) surface by several adsorbed layers of the same molecule

(Dong et al., 2004). Molecular ﬂ uorescence peaks in the spectra became

sharper with increased coverage while the plasmon-related band was

suppressed. Due to the higher thermal drift at room temperature, the

spectra in this observation were averaged over several molecules.

Weakness of interaction between a molecule and the substrate turns

out to be essential for the observation of STM-induced molecular ﬂ uo-

rescence. STM-induced ﬂ uorescence combined with imaging and STS

can be used to probe the interdependence between conformational

structure, energy levels, and optical properties of single molecules. The

inherent stability of the LT-STM together with the suppressed molecu-

lar diffusion at low temperatures should enable the study of electron

dynamics in organic molecules with submolecular resolution.

3.5 Spin-Polarized Scanning Tunneling Microscopy

STM can yield information about magnetic properties by use of spin-

polarized tunneling between a magnetic tip and substrate. The spin

valve effect predicts that the tunneling current depends on the relative

orientation of the magnetic moments of the tunneling electrodes

(Julliere, 1975; Slonczewski, 1988). A magnetic tip acts as a source of

spin-polarized electrons, probing the spin-split density of states of the

magnetic sample. This technique allows imaging with atomic spatial

resolution and, like conventional STM, is mostly sensitive to the topmost

atomic layer. The ability to probe topography, crystallography, and

magnetism at the same time renders spin-polarized scanning tunnel-

ing microscopy (SP-STM) a very powerful tool for the investigation of

magnetic surfaces and monolayers. Since many magnetic phenomena

exist only below a certain critical temperature, it is advantageous to

use a low- or variable-temperature STM to observe magnetic materials.

In addition, measurements on low-dimensional magnetic systems like

ultrathin ﬁ lms or superparamagnetic particles demand low tempera-

tures, since the Curie temperature T

in general scales with dimension.

In addition to those advantages of low temperatures speciﬁ c to SP-

STM, all other advantages mentioned in the introduction apply.

Wiesendanger et al. (1990) explored SP-STM using CrO

tips on an

antiferromagnetic Cr(001) surface. These early experiments repre-

sented a mixture between topographic and spin-dependent contrast.

Discriminating between magnetically caused contrast and contrast

from other features of the electronic density of states near the Fermi

level requires images taken with different tips (magnetic and nonmag-

netic). An experimental setup for an LT-STM, which uses an external

1126 U. Weierstall

magnetic ﬁ eld combined with the ability to rotate the sample without

changing the tip position, has been described (Wittneven et al., 1997).

This setup enables the determination of the relative magnetic orienta-

tion between the end of the tip and the sample.

Other magnetic imaging techniques and their respective lateral reso-

lution are magnetooptical Kerr effect (MOKE) microscopy (∼300 nm),

MOKE with scanning near-ﬁ eld optical microscopy (SNOM) (∼50 nm),

scanning electron microscopy with polarization analysis (SEMPA)

(∼40 nm), magnetic force microscopy (MFM) (∼20 nm), X-ray magnetic

linear dichroism photoelectron emission microscopy (XMLD-PEEM)

(∼20 nm), off-axis electron holography, and Lorentz microscopy for thin

ﬁ lms (∼20 nm). SP-STM has proven to be capable of atomic resolution,

which has been shown for the ﬁ rst time on a manganese monolayer on

W(110) at 16 K (Heinze et al., 2000). A two-dimensional antiferromag-

netic structure as predicted by theory with a periodicity of 4.5 Å

has

been observed as shown in Figure 17–43. This is a considerable advance,

considering that previous characterizations of antiferromagnetic

domains could not go beyond micrometer resolution. Spectroscopic

studies with SP-STM allow the correlation of structural, local elec-

tronic, and local magnetic properties down to the atomic level. This

has been demonstrated in low-temperature studies of ferromagnetic

rare earth (Bode et al., 1998) and transition metal (Pietzsch et al., 2000a)

systems as well as antiferromagnets (Kleiber et al., 2000). The tips used

in these experiments are nonmagnetic tips coated with a thin layer of

ferromagnetic (Kleiber et al., 2000) or antiferromagnetic (Kubetzka

et al., 2002) material. Fe-coated tips are magnetized perpendicular to

the tip axis at the apex and are therefore sensitive to the in-plane com-

ponent of the sample’s magnetization (Bode et al., 1998, Kleiber et al.,

2000). FeGd-coated tips exhibit a perpendicular magnetic anisotropy

and are therefore sensitive to the out-of-plane component of the sam-

ple’s magnetization (Kubetzka et al., 2002).

If the LT-STM is built completely from nonmagnetic materials, SP-

STM can be performed in the presence of high magnetic ﬁ elds, which

enables observation of hysteresis on a nanometer scale (Pietzsch et al.,

2001). The internal spin structure of magnetic vortex states on three-

dimensional Fe islands on W(110) has been resolved recently with an

LT-STM, as shown in Figure 17–44. A Cr-coated tip was used to map

out both the curling in plane magnetization around the vortex core as

well as the perpendicular magnetization within the vortex core (see

Figure 17–45). The width of the core was determined as 9 ± 1 nm in

agreement with theory.

The extreme surface sensitivity of the SP-STM (as of all types of STM)

may be its only weakness with regard to the investigation of coupling

phenomena. Possible changes in the magnetic conﬁ guration of buried

layers therefore are hidden from the measurement. These can be ana-

lyzed with XMLD-PEEM (Scholl, 2003), which, because of its elemental

speciﬁ city and relatively long probing depth (3–5 nm), is able to inves-

tigate layered systems at more modest spatial resolution. However,

SP-STM is unrivaled for the investigation of the magnetic structure of

magnetic monolayers, surfaces, or surface alloys at atomic resolution.

Chapter 17 Low-Temperature Scanning Tunneling Microscopy 1127

Figure 17–43. Atomic

resolution magnetic

imaging with SP-STM.

(A) Constant-current

image of one monolayer

of Mn on W(110)

recorded with a non-

magnetic W tip at 16 K.

(B) Image recorded with

a magnetic Fe tip

showing an antiferro-

magnetic conﬁ guration

as predicted by theory.

The colored insets show

calculated STM images.

theoretical line sections

from (A) and (B). The

image size is 2.7 nm by

2.2 nm. (From Heinze et

al., 2000.) (See color

plate.)

Figure 17–44. Magnetic vortex states on Fe islands imaged with SP-STM. (a) Spin-resolved dI/dV map

of seven monolayers Fe on W(110) measured at 16 K with a Cr-coated tip being sensitive to the in-plane

component of the sample magnetization. The islands exhibit a magnetic vortex state, as visible in (b).

(From Wiesendanger et al., 2004.)

1128 U. Weierstall

Figure 17–45. Magnetic vortex states on Fe islands at imaged higher magniﬁ cation. Spin resolved dI/

dV maps measured with an (a) in-plane and (b) out-of-plane sensitive Cr-coated tip. The curling in

plane magnetization around the vortex core is visible in (a), whereas the magnetization component

in the direction of the surface normal is seen in (b) as a bright area. (c) Spin-resolved dI/dV signal

measured around the circle in (a) indicates that the in-plane component continuously curls around

the core. (d) Spin-resolved dI/dV signal measured along the lines in (a) and (b) showing a vortex core

width of 9 ± 1 nm. (From Wiesendanger et al., 2004.)

Chapter 17 Low-Temperature Scanning Tunneling Microscopy 1129

3.6 Superconductivity

The technique of tunneling was initially used as a tool to understand

the bulk properties of superconductors. Electron tunneling through

thin insulating ﬁ lms into superconductors is used to study how the

density of states in a superconductor is modiﬁ ed. This technique aver-

ages over a macroscopic area and requires a high-quality insulating

tunneling barrier. By using an STM for these measurements, a control-

lable vacuum barrier is used instead and superconductors can be

studied, which do not support the formation of good, high-quality insu-

lating ﬁ lms. Moreover, tunneling takes place over only a small area and

variations in the density of states can be studied on an atomic scale.

According to BCS theory, the superconducting state is a new ground

state where electrons of opposite momentum and spin form pairs. Any

excitation of this ground state requires a Cooper pair to be broken into

two single particle excitations. A minimum energy of ∆, the energy gap,

is required to create one of these quasiparticles. If an electron tunnels

from a metallic STM tip into the superconductor, it will enter into one

of these quasiparticle states. Therefore the tunneling current provides

a direct measure of the density of these states. Superconductors expel

a magnetic ﬁ eld up to a certain critical ﬁ eld (Meissner effect). For larger

ﬁ elds, superconductivity is destroyed. In type II superconductors a

mixed phase exists for certain magnetic ﬁ eld strengths, consisting of

hexagonal lattices of normal conducting vortices (ﬂ ux lines) in the

superconductor, the so-called Abrikosov vortex lattice. Many experi-

ments have conﬁ rmed the existence of these ﬂ ux lines, e.g., electron

microscopy (Bosch et al., 1985) and electron holography (Matsuda et al.,

1989; Harada et al., 1992; Bonevich et al., 1993). These techniques are

sensitive to the magnetic ﬁ eld variations. STM, in contrast, is sensitive

to the electronic structure. The ﬁ rst LT-STM observation of ﬂ ux lines

and the ﬂ ux lattice has been reported by Hess (Hess, 1991; Hess et al.,

1989, 1990a,b, 1991). Surfaces of NbSe

have been studied in a ultralow

temperature STM in a temperature range from 50 mK to 7 K. Differen-

tial conductance spectra in zero applied ﬁ eld showed the development

of the superconducting energy gap below 7.2 K, which opens up to

about 1.1 meV at the lowest temperatures. A second discontinuity was

observed in the density of states at ±34 mV. This feature resulted from

the charge density wave (CDW) gap, since this material also supports

a CDW state. Differential conductance spectra taken at 0 mV from a

single vortex core at an applied ﬁ eld of 500 G showed the normal con-

ducting vortex as a star pattern, which was explained as resulting from

the 6-fold anisotropy of the atomic crystalline band structure. At 0 mV

no state should exist in the superconductor, since this is the middle of

the superconducting gap. Therefore the normal conducting vortex core

was imaged bright on a dark background. At higher ﬁ elds, the vortices

organize into a hexagonal lattice. Figure 17–46 shows this lattice at a

ﬁ eld of 10 kG. The dI/dV map at 1.3 mV was recorded under constant

current feedback conditions with 1.3 mV bias. The question arises

whether the localized current from the STM tip is sufﬁ ciently large to

cause local perturbations such as heating, exceeding the local critical